Reinforcement Learning Virtual School 2021

Explore comprehensive reinforcement learning concepts through this virtual school event organized by ANITI Toulouse in March and April 2021. Dive into fundamental RL principles starting with basic concepts and deep learning foundations, then progress through advanced topics including stochastic bandits, Monte Carlo Tree Search, and Deep Q-Networks with their variants. Master policy gradient methods and actor-critic approaches while learning about their common pitfalls and practical considerations. Discover evolutionary reinforcement learning techniques and their applications in developing more animal-like learning agents. Examine hierarchical reinforcement learning structures and symbolic representations that enhance agent capabilities. Gain practical experience through hands-on sessions using stable-baselines3 and learn essential tips and tricks for implementing RL in real-world scenarios. Study specialized applications including multi-armed bandits in clinical trials, regularized MDPs, model-based RL with regret bounds, micro-data policy search, and efficient motor skills learning in robotics. Benefit from expert insights on reward processing biases in both humans and RL agents, and explore cutting-edge research on leveraging model-learning for extreme generalization across diverse environments.