This is a very interesting core subject for the Master of AI. The first four weeks are dedicated to covering the foundations of reinforcement learning and teaching Deep Q-learning which is used in the projects. From week 5 on, it shifts gears to game theory. From here a range of different game theoretic applications and fields are covered such as population games, repeated games, coalitional game theory, and covers discrete math topics such as social choice and group dynamics. It is very interdisciplinary, and is probably the most "new" content students will be exposed to. The assessment takes the form of a semester long group project, where you build up a simulation from scratch of single agent tabular Q-learning to multi agent deep Q-learning. The project is done in stages throughout the semester. The final stage is quite challenging as it combines both the simulation implementation and game theory into one mega analysis task. However the project only covers content until about week 6, which makes the remaining 6 weeks sort of pointless as you're never assessed on that material. As a result there is less incentive to pay attention to that section of the course in depth. The tutorial staff are very helpful, and the lecturers clearly enjoy teaching the content. Bernd and Julian are great lecturers, clearly very enthusiastic about the content as they explain stuff very well. The content is definitely challenging if you've never encountered this before. Overall enjoyed this subject, though could have been organised better to fully leverage all 12 weeks of content.

Anonymous, Semester 2, 2023