Transfer Learning for Improving Model Predictions in Robotic...
Transcript of Transfer Learning for Improving Model Predictions in Robotic...
![Page 1: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/1.jpg)
Transfer Learning for Improving Model Predictions in Highly Configurable Software
Pooyan Jamshidi, Miguel Velez, Christian KästnerNorbert Siegmund, Prasad Kawthekar
![Page 2: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/2.jpg)
2
![Page 3: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/3.jpg)
Performance Influence of Configuration Parameters
![Page 4: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/4.jpg)
Classic Sensitivity Analysis
Measure
![Page 5: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/5.jpg)
Classic Sensitivity Analysis
Measure Learn
50 - 3*C1 + 20*C2 - 3*C1*C3
![Page 6: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/6.jpg)
Classic Sensitivity Analysis
Measure Learn
50 - 3*C1 + 20*C2 - 3*C1*C3
Reasoning +Adaptation +Debugging +Optimization
![Page 7: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/7.jpg)
Classic Sensitivity Analysis
Measure Learn
50 - 3*C1 + 20*C2 - 3*C1*C3
Reasoning +Adaptation +Debugging +Optimization
High measurement costs
![Page 8: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/8.jpg)
Idea: Transfer Learning
Measure Learn w/TransferLearning
50 - 3*C1 + 20*C2 - 3*C1*C3
Reasoning +Adaptation +Debugging +Optimization
Reuse
![Page 9: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/9.jpg)
Exploiting Similarity
![Page 10: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/10.jpg)
Exploiting Similarity
![Page 11: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/11.jpg)
GP for modeling black box response function
Motivations:1- mean estimates + variance2- all computations are linear algebra3- good estimations when few data
![Page 12: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/12.jpg)
Scenarios and Assumptions
Environment change (configuration option vs. environment change)
Different benchmark/workloadDifferent program versionDifferent hardware
Shape of old and new model similar
![Page 13: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/13.jpg)
Part
Some Results (“it works”)
![Page 14: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/14.jpg)
First Feasibility Demonstration [SEAMS 2017]
Case study & controlled experimentsCan we improve prediction accuracy?Tradeoffs among #source and #target samples?Fast enough?
Subject sys.: Cobot, Apache Storm, Cassandra14
![Page 15: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/15.jpg)
15
SourceModel
TargetModel
Prediction
from 4samples
Prediction w/
TransferLearning
Example:Performance predictionfor CoBot
![Page 16: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/16.jpg)
16
SourceModel
TargetModel
Predictionfrom 4
samples
Prediction w/
TransferLearning
Example:Performance predictionfor CoBot
![Page 17: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/17.jpg)
17
SourceModel
TargetModel
Predictionfrom 4
samples
Prediction w/TransferLearning
Example:Performance predictionfor CoBot
![Page 18: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/18.jpg)
Accuracy andCosts
18
![Page 19: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/19.jpg)
Accuracy andCosts
19
![Page 20: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/20.jpg)
Part
Future work, insights and ideas
![Page 21: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/21.jpg)
Selecting from Multiple Sources (Cost Model)
Source Simulator Target Simulator
Source Robot Target Robot
(1)
(2)
(3)
(4)
(5)
21
![Page 22: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/22.jpg)
Checking Assumptions
How similar are source and target models for real environment changes (workload/infrastructure/code changes)?
Expected similarities:• Constant• Constant / proportional shift• More noise but similar trends• Many features and interactions with similar impact• Many important features and interactions remain important
22
![Page 23: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/23.jpg)
Goal: find best sample points iteratively by gaining knowledge from source and target domain 23
Active learning + transfer learning
Measure Learn w/TransferLearning
50 - 3*C1 + 20*C2 - 3*C1*C3
Reasoning +Adaptation +Debugging +Optimization
Reuse
![Page 24: Transfer Learning for Improving Model Predictions in Robotic …echtzeitsysteme.github.io/fosd2017/themes/yellow-swan/presentatio… · Transfer Learning for Improving Model Predictions](https://reader030.fdocuments.net/reader030/viewer/2022040608/5ec573bf75aebe15262d3f4d/html5/thumbnails/24.jpg)
Transfer Learning for Improving Model Predictions in Highly Configurable Software
Improves the model accuracy up to several orders of magnitude
Is able to trade-off between different number of samples from source and target enabling a cost-aware model
Exploring similarity across environment changes and active learning
24Source Simulator Target Simulator
Source Robot Target Robot
(1)
(2)
(3)
(4)
(5)