Advanced Machine Learning

Course Program

Each year, one of the following programs is taught.

- Tutorial class ( Slides)
  • Introduction to non-smooth optimization; Specific and universal attacks; Transferability; Defense mechanisms
  • First-order algorithms (gradient descent, subgradient descent, proximal algorithm)
  • Fenchel duality
  • Proximal splitting algorithms (forward-backward, Douglas-Rachford, primal-dual, proximal ADMM)

- Lab exercise: Introduction to proximal splitting algorithms ( Jupyter notebook)
  • Feature selection with Lasso
  • Piecewise constant denoising

- Tutorial class ( Slides)
  • Principle; Specific and universal attacks; Transferability; Defense mechanisms
  • Optimization problem
  • Applications

- Lab exercise: Introduction to adversarial ML ( Jupyter notebook)
  • FGSM attack
  • Adversarial retraining

License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License
Creative Commons License