Available Theses Topics
If you are interested in a particular topic listed here for a Bachelor or Master thesis, please contact the corresponding person from the table below.
If you are interested in writing a thesis on another (non-listed) topic within the scope of our group or you want to participate in guided research or an interdisciplinary project, write an email to Mete Ahunbay. Please state your skills and interests and also attach a current CV and a recent grade report. First contact should be established at least one month before registration of the project in order to allow for sufficient time to settle for a suitable topic.
Title | Focus | Contact |
|---|---|---|
| Optimization and Market Design (BSc or MSc thesis) | Various topics | Prof. Martin Bichler |
| Computational Social Choice and Algorithmic Game Theory | Various topics (having passed "Computational Social Choice", "Algorithmic Game Theory", "Markets Algorithms Incentives and Networks" or "Economics & Computation" is required) | Prof. Felix Brandt |
| Maximizing information-gain to improve sample-complexity in games | Tackle an essential question of game theory: "What information is needed to approximately solve a game?" Starting with zero-sum games, maximize information gain to optimally learn about the space of potential solutions. Your research will address the high costs of determining game parameters, aiming to refine game-theoretic reasoning for real-world applications. Requirements: Advanced Python programming skills, algorithmic game theory, information-theory | Fabian Pieroth |
| Learning in Games | Computing Nash Equilibria in complete-information games is known to be PPAD-complete [1]. Thus, determining equilibria in more realistic, Bayesian games is even more challenging and analytically often intractable. However, some learning algorithms emerged in recent years that can approximate equilibria in many of those games [2], [3]. The thesis aims to analyze alternative models that challenge established algorithms. Requirements: good programming skills (pref. in Python and Pytorch), understanding of fundamental game theoretical concepts, good mathematical background [1] C. Daskalakis, P. Goldberg, and C. Papadimitriou. 2009. The Complexity of Computing a Nash Equilibrium. SIAM J. Comput. 39, 1 (Jan. 2009), 195–259. https://doi.org/10.1137/070699652 [2] Bichler, M., Fichtl, M., Heidekrüger, S., Kohring, N., & Sutterer, P. (2021). Learning equilibria in symmetric auction games using artificial neural networks. Nature machine intelligence, 3(8), 687-695. [3] Fichtl, M., Oberlechner, M., & Bichler, M. (2022). Computing Bayes Nash equilibrium strategies in auction games via simultaneous online dual averaging. arXiv preprint arXiv:2208.02036. |
|
| Learning with Contextual Bandits in Auctions | We consider a setting, where bandit algorithms have to learn to bid in repeated auctions. In an extension, where we look at the incomplete-information model [1,2], applying standard bandit algorithms leads to very slow convergence. In this thesis, we want to get an overview over different contextual bandit algorithms [3] (literature research) and run numeric experiments to compare them (implementation). --- Requirements: good programming skills (Python), understanding of fundamental game theoretical concepts, good mathematical backround [1] Bichler, Gupta, Mathews, Oberlechner (2024) - Low Revenue in Display Ad Auctions: Algorithmic Collusion vs. Non-Quasilinear Preferences [link] | Matthias Oberlechner |