13/03 - No homework for lecture 5 and 6.

20/02 - Homework for lecture 4. You can have a look at the third exercise sheet, on optimization for quadratic functions.

13/02 - Homework for lecture 3-4. You can have a look at the second exercise sheet:

• Compulsory exercises to prepare
  • Exercise 1: a different proof for GD (20 min)
  • Exercise 3: another approach to obtain co-coercivity (30 min)
• Recommended exercises to prepare
• Exercise 2: Nesterov AGD (30 min)

13/02 - Correction for ES1 available.

31/01 - No lecture next week on 6/02 ! Next lecture on 13/02.

31/01 - Homework for lecture 2. You can have a look at the first exercise sheet:

• Compulsory exercises to prepare
  • Proof of Ledoux's Lemma (exercise 1) - 20 min
  • Upper bound on the Rademacher complexity (exercise 2) - 15 min
  • Showing that we can apply McDiarmid inequality - 5 min (just check that the condition for McDiarmid's Lemma is valid)
• Optional exercises to prepare
• Zhang's Lemma (exercise 4). Proves the fact that the miminimizer of the expected relaxed risk is also a minimizer of the 0-1 risk - 1h20 min
• Proving McDiarmid inequality (exercise 3) - 30 min

Lecture 4 and 5: Non smooth Stochastic Approximation, Lower bounds.

• Slides from 2022.

Lecture 3

• Exercise sheet number 3.
• Exercise sheet number 3.
• Corresponding slides from 2021-22.

Lecture 2

• Exercise sheet number 2.
• Exercise sheet number 2 - solution.
• Corresponding slides from 2021-22.

Lecture 1

• Exercise sheet number 1.
• Exercise sheet number 1 - solution.
• Corresponding slides from 2021-22.

Ressources

• (old) slides 2021-2022 .
• Book draft by Francis (contains all content and more !) .

Dates

• 30 Jan: Lecture 1
• 6 Feb: no lecture
• 13 Feb: Lecture 2
• 20 Feb: Lecture 3
• 27 Feb: vacances
• 6 Mar: Lecture 4
• 13 Mar: Lecture 5
• 20 Mar: Lecture 6
• 27 Mar: Final Exam

• Registration link ! Please register asap !!!