Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration...

69
© 2018 The MathWorks, Inc. © 2015 The MathWorks, Inc. Are you ready for AI? Is AI ready for you? Dr. Mischa Kim

Transcript of Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration...

Page 1: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inccopy 2015 The MathWorks Inc

Are you ready for AI

Is AI ready for you

Dr Mischa Kim

copy 2018 The MathWorks Inc

Alexa ndash

Write my Expo

keynote for me

copy 2018 The MathWorks Inc

Alexa ndash

Play Mozart 119136119135119136

copy 2018 The MathWorks Inc

Source Gartner Real Truth of Artificial Intelligence by Whit Andrews

Presented at Gartner Data amp Analytics Summit 2018 March 2018

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 2: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Alexa ndash

Write my Expo

keynote for me

copy 2018 The MathWorks Inc

Alexa ndash

Play Mozart 119136119135119136

copy 2018 The MathWorks Inc

Source Gartner Real Truth of Artificial Intelligence by Whit Andrews

Presented at Gartner Data amp Analytics Summit 2018 March 2018

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 3: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Alexa ndash

Play Mozart 119136119135119136

copy 2018 The MathWorks Inc

Source Gartner Real Truth of Artificial Intelligence by Whit Andrews

Presented at Gartner Data amp Analytics Summit 2018 March 2018

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 4: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Source Gartner Real Truth of Artificial Intelligence by Whit Andrews

Presented at Gartner Data amp Analytics Summit 2018 March 2018

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 5: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 6: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 7: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to imitate

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 8: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Artificial Intelligence

The capability of a machine to match or exceed

intelligent human behavior

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 9: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Artificial Intelligence Today

The capability of a machine to match or exceed

intelligent human behavior

by training a machine to learn the desired behavior

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 10: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Traditional Programming

COMPUTER

Program

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 11: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERProgram

Output

Data

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 12: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

There are two ways to get a computer to do what you want

Machine Learning

COMPUTERModel

Output

Data

Artificial Intelligence Machine Learning

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 13: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 14: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 15: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 16: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 17: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

EVERYTHING

ELSE

Data

Model

Output

Access Data

Analyze Data

Develop

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 18: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

DevelopAnalyze DataAccess Data Deploy

Data

Model

Output

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 19: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Develop

AI model

Modeling amp

simulation

Algorithm

development

Are you ready for AI

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Deploy

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 20: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 21: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI

Develop

Data

exploration

Preprocessing

Analyze Data

Domain-specific

algorithms

Sensors

Files

Access Data

Databases

Desktop apps

Enterprise

systems

Deploy

Embedded

devices

AI model

Modeling amp

simulation

Algorithm

development

Caffe

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 22: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Do you need AI

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 23: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 24: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each robot

bull Predict and fix failures before they happen

bull AI handles uncertainty and variability

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 25: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 26: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 27: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

What is crispiness

Crispy

Crushing Sound

+Crispy Enough

Soggy

=Crushing Force

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 28: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 29: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Replicating human perception with machine learning

Technical University of Munich

Classification Learner

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 30: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Fresh

Soggy

SoggyFresh

Tru

e C

las

s

Predicted Class

93

91

91

91

89

95

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 31: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if yoursquove never used machine learning

No experience required

Use apps to try out all possible models

Use domain expertise and familiar tools to prepare data

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 32: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Find out more

ldquoMATLAB을이용한머신러닝 (기본)rdquo

ldquo복잡한문제를단순하게만드는MATLAB환경에서의머신러닝 (중급)rdquo

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 33: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

You canrsquot identify features in your data

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 34: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 35: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Use deep learning to identify features automatically

Feature extraction Classification

Crispy

Crispy enough

Soggy

Data

Machine Learning Workflow

Crispy

Crispy enough

Soggy

Deep neural network

95

3

2

hellip

Data

Deep Learning Workflow

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 36: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 37: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 38: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Traditional Approach

bull Geologists assess seven different metrics

bull Can take hours to analyze one site

bull Critical shortage of geologists

New Approach

bull Use deep learning to automatically

recognize metrics based on images

bull On-site evaluators decide with support

from deep learning

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 39: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

Label each

sub-image

Split into

sub-images

Image

Weathering

Alteration

(1-4)

Fracture

Spacing

(1-5)

Fracture

State

(1-5)

3 3 2

4 1 1

2 3 2

3 3 2

⋮ ⋮ ⋮ ⋮

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 40: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 41: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

MATLAB Production Server

Efficient tunnel drilling with deep learning

Obayashi Corporation

AlexNetPRETRAINED MODEL

Ice cream Teapot Goose

Transfer learning

Custom NetworkWeathering alteration 4

Fracture spacing 3

Fracture state 2

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 42: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learningDeep learning in 5 lines of code

nnet = alexnet

cam = webcam

picture = snapshot(cam)

picture = imresize(picture[227 227])

label = classify(nnet picture)

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 43: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Deep learning in 5 lines of code

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 44: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 45: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if you canrsquot identify features in your data

Deep learning

Transfer learning

Automation and AI to label data

Point cloud semantic segmentation

Classification

Car

Truck

Ground

Background

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 46: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Find out moreldquo개발에서 구현까지 MATLAB 환경에서의딥러닝rdquo

ldquo딥러닝을 활용한 영상 인식 응용프로그램개발 워크플로우rdquo

ldquoGPU 기반의 임베디드 하드웨어에서의 딥러닝및 코드 생성

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 47: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

If you donrsquot have the right data

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 48: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

AI for Predictive Maintenance

bull Measure the wear of each blade

bull Predict and fix failures before they happen

bull Canrsquot rely on failures in the field

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 49: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Simulink model

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 50: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Predictive maintenance with synthetic failure data with

MATLAB amp Simulink

Measured data

Inject failures

Refine model

Simulink model

Failure data

Failure conditions

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 51: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if you donrsquot have the right data

Generate data with simulations

Simulation environment for

reinforcement learning

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 52: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Find out more

ldquo고장 진단 및 건전성 관리 기술 알고리즘개발에서 사물인터넷 환경 구성까지rdquo

ldquo물리모델 시뮬레이션을 활용한 고장 예측rdquo

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 53: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 54: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Low-carbon homes

bull Generate power with fuel cell and solar panels

bull Store power in battery

bull Buy power when needed sell when extra

bull Record data on environment and energy usage

Goals

bull Minimize energy cost

bull Use EV battery for additional storage

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 55: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 56: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

HomeHome Energy Controller

Battery

command

Generated and consumed power

Stored energy

Electricity prices

Predicted

vehicle use

yen

Simscape Power SystemsModel predictive control

Mixed integer linear programming

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 57: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Parallel

computing

1000 CSV Files

Access Data Deploy

PreprocessingClassification

Learner

Analyze Data

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 58: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 59: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Optimizing home energy management system

Denso

Develop

Data

Simulink

Control

algorithms

Parallel

computing

1000 CSV Files

Access Data Deploy

Optimization

Preprocessing Embedded

devices

Classification

Learner

Analyze Data

Simscape Power

Systems

ldquoThe effort would have taken significantly longer if we

had used disparate tools

[MATLAB] enabled our team of domain experts who

lacked formal training in data science machine learning

and parallel computing to incorporate all these areas in

our design processrdquo

Akira Ito and Ryu Matsumoto

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 60: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 61: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 62: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Autonomous

Primary

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 63: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 64: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

++

Host computer ProsthesisProcessing laptop

Microphone

EMG

Drummer

Music

Exceeding human capabilities with a robotic drumming prosthesis

Georgia Tech Center for Music Technology

PID controller

AI algorithms

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 65: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 66: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning Easy programming

Apps

Domain expertise to prepare data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 67: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 68: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

Are you ready for AI if hellip

Yoursquove never used machine learning

You canrsquot identify features in your data

You donrsquot have the right data

Easy programming

Apps

Domain expertise to prepare data

Deep learning identifies features for you

Transfer learning works with less data

Use AI to label data

Generate failure data with simulations

Simulate environment for reinforcement learning

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI

Page 69: Are you ready for AI? Is AI ready for you? · Are you ready for AI? Develop Data exploration Preprocessing Analyze Data Domain-specific algorithms Sensors Files Access Data Databases

copy 2018 The MathWorks Inc

With MATLAB and Simulink you ARE ready for AI