Reinforcement Learning section of the Algorithms in Machine Learning class at ISAE-Supaero
This class covers an introduction to Reinforcement Learning (RL) in 18 hours, over 6 sessions. It aims to provide both a solid theoretical foundation and a quick learning curve towards current Deep RL algorithms. It starts with the fundamental notions underlying RL: Markov Decision Processes, model-based resolution approaches including Dynamic Programming, sample-based resolution of the Bellman equation. This leads to the identification of the three bottomline challenges in RL: function approximation, the exploration/exploitation trade-off and the search for optimality. This provides perspective to the following classes that introduce methods designed to tackle these challenges, including Deep RL methods. By the end of the class, students should be able to understand the literature on RL, implement key algorithms, and anticipate the difficulties of applying RL to various problems.
The class is split into a series of notebooks that serve as lecture material, textbook and exercice book.
Open the notebooks in Binder:
Great books available online:
Reinforcement Learning, an introduction
Algorithms for Reinforcement Learning
An introduction to Deep Reinforcement Learning
FAQ on installing Gym for Mac users
| Schedule | |||
|---|---|---|---|
| MDPs and their resolution | 08h30 - 11h45 | 03/02/2021 | RL intuitions, Markov Decision Processes, Dynamic Programming |
| Sample-based policy search | 08h30 - 11h45 | 09/02/2021 | Formulations of RL algorithms, Temporal Differences, Q-learning, the 3 bottlenecks of RL |
| Value function approximation | 13h00 - 16h15 | 09/02/2021 | Linear approximations, Deep Q-Networks |
| Policy gradients | 09h00 - 12h15 | 15/02/2021 | PG and Deep PG methods |
| MCTS | 09h00 - 12h15 | 17/02/2021 | Monte Carlo Tree Search |
| open | 13h45 - 17h00 | 17/02/2021 | open session on an RL challenge |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent