MSc Seminar Reinforcement Learning (Dereich)

Do 10-12

Prof. Dr. Steffen Dereich

Topic:

Seminar on Reinforcement Learning

Reinforcement learning is highly relevant in developing intelligent systems that learn optimal behaviors through trial and error, making it crucial for applications like robotics, game AI and autonomous decision-making. 

In this seminar we focus on Nash equilibria and their role in reinforcement learning (RL). We start with an introduction to fundamental concepts, including the definition and significance of Nash equilibria in strategic decision-making and an overview of reinforcement learning techniques. Then we want to discuss current articles in this field.

No prior expertise in game theory or reinforcement learning is required.

This course can be counted as either the first or second part of the module “Ergänzungen und Wissenschaftliches Arbeiten” and students have two options:

Option 1: The student participates In the "regular" seminar, where we study topics around the ergodic theory of Markov processes following the lecture notes of Martin Hairer. In this case, the student will give a presentation, which will be graded and can be counted as the first or second part of the abovementioned module.

Option 2: The student participates in the seminar as preparation for a Masters thesis. For this, the student contacts one of the three lecturers regarding possible master’s thesis topics. In consultation with the lecturer, the student will work on the corresponding topic. Towards the end of the semester, the students present their preliminary work, for example, by presenting a relevant scientific article. This performance is ungraded and can be counted as the second part of the module.