Towards Performatively Stable Equilibria in Decision-Dependent Games for Arbitrary Data Distribution Maps
The episode discusses a paper titled "Towards Performatively Stable Equilibria in Decision-Dependent Games for Arbitrary Data Distribution Maps." The hosts explain that this research finds stable points in decision-dependent games where players optimize based on shifting data. The paper introduces a measurable sensitivity measure to replace unknown smoothness constants, proving convergence even with finite samples and outperforming existing methods in real-world game simulations.