Week 4: Reinforcement Learning

[pdf slides]

Carl Henrik Ek

Abstract:

In this we’ll introduce Reinforcement Learning (RL) as a natural extension of probabilistic modeling and statistical inference. We can think of RL as a specific recursive probabilistic model designed for scenarios where simultaneous data acquisition is infeasible. We’ll cover fundamental concepts including Markov Decision Processes, the distinction between model-based and model-free approaches, and Q-learning. This will be related to the challenge of data acquisition and the balance between exploration and exploitation.