ANALYZING REINFORCEMENT LEARNING ALGORITHMS USING EVOLUTIONARY GAME THEORY

The cover image visualizes several learning trajectories of Lenient Frequency Adjusted Q-learning in the Matching Pennies game, together with the dynamics of its evolutionary model. This image is related to the experiments described in Section 6.1. Acknowledgments A master’s thesis is not something that suddenly springs into existence. Not completely unrelated to my research, I came to think of it as a dynamical process, sometimes slow moving, sometimes fast, in one direction or another. Only if you know the governing dynamics, will you know what the end result will be. To extend this metaphor further (again not completely unrelated to my research), thesis writing resembles multi-agent interactions, where you are not on your own, but influenced by a group of people around you. Without these influences, this thesis would not have been anything like it is now. First of all, I would like to express my gratitude to my two supervisors. Karl, thank you for your support, enthusiasm, and constructive comments during our many meetings. There have been periods that my research did not proceed as well as I hoped, or that I lacked motivation or inspiration; but every time one meeting with you was enough to get me back on track. I hope that we get the chance to continue this fruitful cooperation in the future. Michael, thank you for all your good comments and suggestions. I would also like to thank all my good friends and fellow students at the Department of Knowledge Engi-neering, for making my time there a very enjoyable one. Furthermore, I thank everyone who supported me in any way during my study, especially my closest friends and family. Without you I would not have been where I am now. As an old saying goes, “the proof of the pudding is in the eating”. Therefore, I hope that you will enjoy reading this thesis as much as I enjoyed the process of writing it