Slides: see slides folder:

(1) Background

• MLTutorialSPLC20-SLR.pdf
• MLTutorialSPLC20-ML.pdf
• DT.pdf

(2) The case of VaryLaTeX

• MLTutorialSPLC19-VaryLaTeX.pdf
• MLTutorialSPLC19-VaryLaTeXExercice.pdf

(3) The case of x264

• MLTutorialSPLC20-ICPE_Sampling.pdf
• MLTutorialSPLC19-x264Exercice.pdf

(4) Conclusion

• MLTutorialSPLC20-WrapUp.pdf

Jupyter notebook: see latex and x264 folder:

• The case of VaryLaTeX: Latex_specialization.ipynb
• The case of x264: x264_performance_prediction.ipynb

Videos:

• https://www.youtube.com/watch?v=u1ralqbHCyM
• https://www.youtube.com/watch?v=0ePwyCCJyIM
• https://www.youtube.com/watch?v=msyrM1FJXs0
• https://www.youtube.com/watch?v=eyNE_K3eCE8
• https://www.youtube.com/watch?v=jtXoYdHhVtI

• Welcome and general motivation: Why machine learning is relevant for engineering software configurable systems?
• The VaryLaTeX case (demonstration)
  • https://github.com/diverse-project/varylatex
  • youtube videos: https://www.youtube.com/playlist?list=PLcDsXHkK7hJ3n9v7VAMbZCIV6egreI8OL
  • see MLTutorialSPLC19-VaryLaTeX.pdf
• Overview: An overview of works in the field, see MLTutorialSPLC19-SLR.pdf
  • based on a systematic literature survey
  • we describe the different applications (pure prediction, optimization, specialization, understanding, etc.)
  • we review subject systems and application domains
  • we describe numerous sampling strategies
  • we detail how configurations are measured
  • we report on learning algorithms used and their assessment
• Setup instructions (1020 => 1030)
• Practical session 1: learning-based specialization with VaryLaTeX case (1100 => 1145, see MLTutorialSPLC19-VaryLaTeXExercie.pdf)
• information about VaryLaTeX: https://hal.inria.fr/hal-01659161/ https://github.com/FAMILIAR-project/varylatex
• dataset: https://github.com/FAMILIAR-project/varylatex/blob/master/output-FSE/csvs/
• decision tree algorithm and a focus on interpretability
• we use Python and Jupyter notebooks
  • R solution is also available https://github.com/FAMILIAR-project/varylatex/blob/master/output-FSE/analysisCT.R for the interested audience
• exercices:
  • change the training set size and analyze the effect on accuracy and rules
  • change some hyperparameters
  • change the algorithm (using random forest)
• Practical session 2: performance prediction with x264 case (1145 => 1215, see MLTutorialSPLC19-x264Exercice.pdf)
• dataset from the literature
• we use Python and Jupyter notebooks
• Summary and open research directions (1215 => 1230, see MLTutorialSPLC19-WrapUp.pdf)
  • wrap-up
  • open issues

requirements: Jupyter, Python 3 with scikit-learn, pandas, and numpy play with notebooks in latex (for VaryLaTeX exercice) and x264 (for x264 exercice)

pip install pandas
pip install numpy
pip install scikit-learn
pip install jupyter
pip install graphviz

Docker alternative :

docker build -t splc .

docker run -p 8888:8888 splc