CNRS Junior Researcher at CRIStAL

emilie.kaufmann"at"univ-lille.fr

Inria Lille - Nord Europe

Equipe SequeL, Bureau A07

40, avenue du Halley

59650 Villeneuve d'Ascq, FRANCE

+333.59.57.79.12.

This 24 hours Reinforcement Learning class is jointly taught with Omar Darwiche-Domingues (Inria SequeL) and Olivier Pietquin (Google Brain). The slides of the course and notebooks for the practical sessions will be made online the day before the class:

- Lecture 0: Introduction to Reinforcement Learning.
- Lecture 1: Markov Decision Processes. Numerical illustrations: Notebook 1. Notebook 2.
- Lecture 2: Multi-armed bandits.
- Practical session 1: Bandit algorithms. Code.Correction.
- Lecture 3: Solving a MDP with Known Parameters.
- Practical session 2: Value Iteration and Policy Iteration. Code.
- Lecture 4: First Reinforcement Learning Algorithms.
- Lecture 5: Reinforcement Learning with Approximation.
- Lecture 6 and 7: (by Olivier Pietquin) Scaling Up Reinforcement Learning.
- Practical session 3: Implementing DQN. Code.
- Practical session 4: Implementing Policy Gradient algorithms. Code.
- Lecture 8: Bandit tools for Reinforcement Learning.

The course will be illustated by 4 practical sessions in Python, using jupyter notebook. To validate the class, you need to do a project (report + oral presentation in front of the class on Thursday). UPDATE: project presentation morning is February 13th, 2019 from 8h-12h15. Deadline for the report is February 7th, 11pm.

A list of possible projects is available here, but you can also propose something else. Once you have chosen your project, please register here.

- Reinforcement Learning. Richard Sutton and Andrew Barto (new 2018 edition). The book is available online here.
- Reinforcement Learning Algorithms. Csaba Szepesvari (2009). The book is online here.
- Neuro-Dynamic Programing. Dimitri Bertsekas and John Tsitsiklis (1996).
- Markov Decision Processes. Martin Puterman (1994).
- Bandit Problems. Tor Lattimore and Csaba Szepesvari (2019). The book is available online here.
- Nice lecture notes written by several colleagues: Rémi Munos, Alessandro Lazaric and Aurélien Garivier.
- Material from the 1st Reinforcement Learning Summer School.