Reinforcement learning and its implementation in Julia

Lecturer: 

Clemens Heitzinger, TU Wien (Technical University Vienna)

Schedule and place:

When: This 15-hour course will take place in 5 days, with morning/afternoon sessions of 1.5 hours, November 18-22, 2024

Where: UCLouvain - Euler building (room A.002) Avenue Georges Lemaître,4 - 1348 Louvain la Neuve.

and Maxwell building - Shannon room (a.105) for 19 November 13:30-15:30

Travel instructions are available here

Planned Schedule: 

09:45 - 10:15 Welcome coffee

10:15 - 11:45

13:30 - 15:00

Abstract: 

Reinforcement learning is the subfield of machine learning that has enabled the advances in artificial intelligence in the last couple of years, including ChatGPT as well as solving chess, Go, and Atari 2600 games by a single algorithm.

This course provides an introduction to the theory of reinforcement learning first, followed by a discussion of the recent developments and current algorithms. Standard environments (problem statements) for programming exercises will be provided as well.

+ Markov decision processes
+ Standard environments / problem statements Dynamic programming Monte
+ Carlo Temporal-difference methods incl. Q-learning Deep reinforcement
+ learning Distributional reinforcement learning

Course material:

The course will be based on the books: Reinforcement Learning: Algorithms and Convergence and Algorithms with Julia

Evaluation:

TBD in the first lecture