What’s New in MATLAB and Simulink · Using MATLAB and Simulink to Build Deep Learning Models...

55
© 2019 The MathWorks, Inc. 1 © 2015 The MathWorks, Inc. What’s New in MATLAB and Simulink 山本 順久

Transcript of What’s New in MATLAB and Simulink · Using MATLAB and Simulink to Build Deep Learning Models...

© 2019 The MathWorks, Inc. 1© 2015 The MathWorks, Inc.

What’s New in MATLAB

and Simulink

山本順久

© 2019 The MathWorks, Inc. 2

© 2019 The MathWorks, Inc. 3

Algorithms in Everything

© 2019 The MathWorks, Inc. 4

Using MATLAB & Simulink to Build Algorithms in Everything

Inputs Design OutputsArchitecture

Spanning

Test & Verification Collaboration Scaling

© 2019 The MathWorks, Inc. 5

The capability of a machine to

match or exceed intelligent human behavior by

training a machine

to learn the desired behavior

Artificial Intelligence

training a machine

to learn the desired behavior

© 2019 The MathWorks, Inc. 6

Output

Program

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

Traditional Programming

COMPUTER

Data

© 2019 The MathWorks, Inc. 7

Program

Output

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

Machine Learning

COMPUTER

Data

Model

© 2019 The MathWorks, Inc. 8

Input

Using MATLAB and Simulink to Build Deep Learning Models

Inputs

Input

Design

Design

Outputs

Outputs

Data

Machine

Learning

Deep

Learning

Model

© 2019 The MathWorks, Inc. 9

Using Apps for Ground Truth LabelingImage and Video Data

InputDesi

gn

Outp

uts

Computer Vision Toolbox

Input

© 2019 The MathWorks, Inc. 10

Using Apps for Ground Truth LabelingSignal Data

Input

Signal Processing Toolbox

Input

© 2019 The MathWorks, Inc. 11

Using Apps for Ground Truth LabelingAudio Data

Input 2 3

Audio Toolbox

Input

© 2019 The MathWorks, Inc. 12

Using Apps for Designing Deep Learning Networks 2 Design

Deep Learning Toolbox

3

© 2019 The MathWorks, Inc. 13

Using Models from Other Frameworks

(…)Caffe

Keras-

Tensorflow

PyTorch

Open Neural Network Exchange

MATLAB ONNX

CNTK

Caffe2

MXNet

Core ML

Deep Learning Toolbox

2 Design 3

© 2019 The MathWorks, Inc. 14

OutputDeploying Deep Learning Applications

ARM

Compute

Library

Intel

MKL-DNN

Library

Design Output

Deep

Learning

Application

Coder

Products

NVIDIA

TensorRT &

cuDNN

Libraries

MATLAB CoderGPU Coder

Inputs

Pre-

processing

Post-

processing

Input Output

© 2019 The MathWorks, Inc. 15

OutputDeploying Deep Learning Applications

ARM

Compute

Library

Intel

MKL-DNN

Library

Design Output

Deep

Learning

Application

Coder

Products

NVIDIA

TensorRT &

cuDNN

Libraries

MATLAB CoderGPU Coder

Inputs

Pre-

processing

Post-

processing

Input Output

Find out more:

ディープラーニングアプリケーションの組み込みGPU/CPU実装

MathWorks Japan 町田和也

B4 16:30–17:00

© 2019 The MathWorks, Inc. 16

Using MATLAB and Simulink for Reinforcement Learning

Reinforcement Learning Toolbox

© 2019 The MathWorks, Inc. 17

Output

Using MATLAB and Simulink to Build Deep Learning Models

Inputs

Input

Design

Design

Outputs

Output

Data

Machine

Learning

Deep

Learning

Model

Using MATLAB and Simulink for Reinforcement Learning

Reinforcement Learning Toolbox

Find out more:

強化学習:最適制御のためのディープラーニングの応用

MathWorks Japan 吉田剛士

B2 14:30-15:00

© 2019 The MathWorks, Inc. 18

Input

Using MATLAB & Simulink to Build Algorithms in Everything

Inputs

Input

Design

Design

Outputs

Outputs

© 2019 The MathWorks, Inc. 19

Creating Your Own Data 2

Simulink

InputInput

© 2019 The MathWorks, Inc. 20

2InputInput

© 2019 The MathWorks, Inc. 21

Identifying the Useful Data 2 Design

Acquire Data

Identify

Condition

Indicators

Deploy &

IntegrateTrain Model

Preprocess

Data

Visualize data

Extract Features

Select the most

useful features

Machine

Learning

© 2019 The MathWorks, Inc. 22

Identifying the Useful Data 2 Design

Predictive Maintenance Toolbox

© 2019 The MathWorks, Inc. 23

Identifying the Useful Data 2 Design

Predictive Maintenance Toolbox

© 2019 The MathWorks, Inc. 24

Identifying the Useful Data

Predictive Maintenance Toolbox

2 Design

© 2019 The MathWorks, Inc. 25

Identifying the Useful Data

Predictive Maintenance Toolbox

2 Design

Find out more:

故障予測xIoT

~予知保全システムの構築~

MathWorks Japan王暁星

C3 15:30-16:00

© 2019 The MathWorks, Inc. 26

Designing Decision Logic with Stateflow

StateflowMATLAB

in MATLAB 2 Design

© 2019 The MathWorks, Inc. 27

Using Stateflow in MATLAB

StateflowMATLAB

2 Design

© 2019 The MathWorks, Inc. 28

Editing at the Speed of Thought

Simulink

2 Design

© 2019 The MathWorks, Inc. 29

Editing at the Speed of Thought

Simulink

2 Design

© 2019 The MathWorks, Inc. 30

Viewing Generated Code Alongside the Model

Embedded Coder

2 OutputOutput

© 2019 The MathWorks, Inc. 31

Sharing Live Scripts

MATLAB

2 OutputOutput

© 2019 The MathWorks, Inc. 32

Sharing Live Scripts

MATLAB

2 OutputOutput

© 2019 The MathWorks, Inc. 33

Arch

Design OutputInput Arch

Evaluating Architectures

Inputs Design OutputsArchitecture

© 2019 The MathWorks, Inc. 34

Designing System and Software Architectures

System1Composer1

Design OutputInput

ArchArch

© 2019 The MathWorks, Inc. 35

Designing System and Software Architectures

System1Composer1

ArchArch

© 2019 The MathWorks, Inc. 36

Designing System and Software Architectures

System1Composer1

ArchArch

Find out more:

Simulink Requirementsと新製品System Composerによるシステムズエンジニアリング

MathWorks Japan 大越亮二

F2 14:30-15:00

© 2019 The MathWorks, Inc. 37

Designing Beyond System and

Software Architectures

AUTOSAR SoftwareSystems and

SoftwareSoC Hardware

and Software

System Composer

SoC Blockset

AUTOSAR Blockset

Design OutputInput

ArchArch

© 2019 The MathWorks, Inc. 38

ArchArchitecture

OutputDesignArchInput

Using MATLAB & Simulink to Build Algorithms in Everything

Inputs Design Outputs

Spanning

Test & Verification Collaboration Scaling

© 2019 The MathWorks, Inc. 39

Architecture

OutputDesignArchInput

Using MATLAB & Simulink to Build Algorithms in Everything

Inputs Design Outputs

Test & Verification

Test & Verification Collaboration Scaling

© 2019 The MathWorks, Inc. 40

Using the MATLAB App Testing Framework

testCase.press(myApp.checkbox)

testCase.choose(myApp.discreteKnob, "Medium")

testCase.drag(myApp.continuousKnob, 10, 90)

testCase.type(myApp.editfield, myTextVar)

MATLAB

Design OutputInput Arch

Test & VerificationTest & Verification

© 2019 The MathWorks, Inc. 41

Authoring Logical Tests for Simulink Models

Simulink Test

Textual Requirements Format

Temporal Assessments

Wireless Test

Observers

Test & Verification

Design OutputInput Arch

Test & Verification

© 2019 The MathWorks, Inc. 42

Using Continuous Integration

MATLAB https://plugins.jenkins.io/

Design OutputInput Arch

Test & VerificationTest & Verification

© 2019 The MathWorks, Inc. 43

Integrating with Third-party Requirements Tools

Simulink Requirements

SimulinkRequirements

External Requirements

Authored Requirements

External Requirements

Requirements

Management

Tools

ReqIF

Import

Edit

Export

Test & Verification

Design OutputInput Arch

Test & Verification

© 2019 The MathWorks, Inc. 44

Using Projects in MATLAB

MATLAB

Design OutputInput Arch

Test & VerificationCollaboration

© 2019 The MathWorks, Inc. 45

Scaling Computations on Clusters and Clouds

MATLAB Parallel Server

MATLAB

Parallel Computing Toolbox

Cloud

GPU

Multi-core CPU

MATLAB Parallel Server

(MATLAB Distributed

Computing Server)

Design OutputInput Arch

ScalingScaling

© 2019 The MathWorks, Inc. 46

Scaling

Input Design OutputArch

Using MATLAB & Simulink to Build Algorithms in Everything

Inputs Design OutputsArchitecture

Scaling

Test & Verification Collaboration Scaling

© 2019 The MathWorks, Inc. 47

Specialized Tools for Building Algorithms

in Everything

Analog Mixed-SignalCommunications Physical interconnects

5G Toolbox

SerDes Toolbox

Mixed-Signal Blockset

Scaling

Design OutputInput Arch

© 2019 The MathWorks, Inc. 48

Developing Autonomous Systems

Perception

Planning

Control

© 2019 The MathWorks, Inc. 49

Evaluate Sensor Fusion Architectures

Sensor Fusion and Tracking Toolbox

© 2019 The MathWorks, Inc. 50

Design Lane-following and Spacing

Control Algorithms

Model Predictive Control Toolbox

© 2019 The MathWorks, Inc. 51

Developing Autonomous Systems

UAV AlgorithmsLidar Processing

& Tracking

Computer Vision Toolbox

HERE HD Maps &

OpenDRIVE Roads

Automated Driving Toolbox

Robotics System Toolbox

© 2019 The MathWorks, Inc. 52

Developing Autonomous Systems

UAV AlgorithmsLidar Processing

& Tracking

Computer Vision Toolbox

HERE HD Maps &

OpenDRIVE Roads

Automated Driving Toolbox

Robotics System Toolbox

Find out more:

ADAS・自動運転アルゴリズム検証のためのシナリオ生成とシミュレーション

MathWorks Japan 大塚慶太郎

D4 16:30-17:00

© 2019 The MathWorks, Inc. 53

Using MATLAB & Simulink to Build Algorithms in Everything

Inputs Design OutputsArchitecture

Spanning

Test & Verification Collaboration Scaling

© 2019 The MathWorks, Inc. 54

Get Started

Deep Learning Onramp

Learn to use deep learning techniques

in MATLAB for image recognition.

MATLAB Onramp

Quickly learn the

essentials of MATLAB.

Simulink Onramp

Learn to create, edit, and

troubleshoot Simulink models.

© 2019 The MathWorks, Inc. 55© 2015 The MathWorks, Inc.