M2 Reinforcement Learning

Reinforcement Learning (RL) is one of the main three subfields of Machine Learning (the other two being Supervised and Unsupervised Learning).

RL considers an agent interacting with an environment. The goal is  to find the policy (i.e., which action to select in which state) yielding optimal cumulative results depending on the time horizon, and the instant rewards given by the environment to the agent's actions.

Examples of RL problems are robotics and games (the game of Go will be used to illustrate some of the policies).

Another part of the course will be devoted to Causal Modelling: identifying causal models and finding the effects of interventions on the world.