Date: October 6, 2025
Time: 3:45-5 p.m.
Location: Gates Hall 310 or via Zoom
Speaker: Natalie Collina
Title: Menus: A Framework for Learning Against Strategic Opponents
Abstract: Menus provide a geometric framework for reasoning about learning and equilibrium in games. A menu captures the set of outcomes a learning algorithm makes available to its opponents, turning questions about long-run play into questions about convex structure. I will introduce the menu framework and focus on how it leads to new insights in extending swap regret and correlated equilibria beyond normal-form games, where menus characterize when algorithms are manipulable and motivate new regret notions that remain tractable in complex games. I will also briefly describe how the same framework underlies results on Pareto-optimal learning algorithms and on optimizing against unknown opponents, highlighting menus as a unifying tool across problems in learning and interaction.
Related papers:
Pareto-Optimal Algorithms for Learning in Games — arXiv:2402.09549
Learning to Play Against Unknown Opponents — arXiv:2412.18297
Swap Regret and Correlated Equilibria Beyond Normal-Form Games — arXiv:2502.20229
Bio: Natalie is a Ph.D. student (2021-present) in Computer Science at the University of Pennsylvania, where she is advised by Michael Kearns and Aaron Roth. Natalie works on problems at the intersection of AI and Game Theory. She is especially interested in understanding repeated strategic interactions between human and AI agents. Her work has been awarded with an IBM PhD Fellowship in Trustworthy AI and a joint Best Paper Award/Best Student Paper Award from the ACM Conference for Economics and Computation.