Note: due to the COVID-19 pandemic, the course will be run fully online using zoom (lecture link TBA) and OLAT (Link). If you would like to attend the lecture and exercise sessions please register in OLAT or contact Jochen Triesch.

Overview

Reinforcement Learning is a research field at the intersection of Computer Science/Engineering (artificial intelligence, machine learning, control theory, robotics) and Neuroscience/Psychology (human and animal learning, reward systems, motivation). It describes how agents (biological or artificial) can learn to optimize their behavior in the presence of feedback that takes the form of rewards and punishments, or how they can learn driven by their own curiosity. This course provides an introduction to the theory of reinforcement learning and discusses applications of these concepts to artificial intelligence and modeling learning processes in biological systems. Covered topics include: Markov Decision Processes, Dynamic Programming, Monte Carlo Methods, Temporal Difference Learning, Biology of reward systems, Plannning, Function Approximation, Deep Reinforcement Learning, Intrinsic Motivation, Hierarchical Reinforcement Learning, Multi-Agent Reinforcement Learning. 

Textbook

The course is based on the text "Reinforcement Learning" by Rich Sutton and Andrew Barto, MIT Press, 2018.

Useful Links

• Link to OLAT Course: Organisation of the course and slides
• Link to Zoom lecture: TBA
• The book: Reinforcement Learning, an Introduction by Richard S. Sutton and Andrew G. Barto
• Scholarpedia has articles on a number of relevant topics.