Core Concepts
Learning to estimate unknown aggregator for robust Nash equilibrium.
Abstract
In this study, the authors investigate the robustness of Nash equilibria in multi-player aggregative games with coupling constraints. They propose an inverse variational inequality-based relationship to estimate the unknown aggregator, allowing for the computation of robust Nash equilibria. The study focuses on scenarios where players' weight in the aggregator is unknown, treating it as a "black box." By disassembling the black-box aggregator and estimating players' weights, they achieve robust Nash equilibrium. The learning approach involves recovering the black-box aggregator from data and reconstructing the problem from a robustness perspective. Simulation experiments demonstrate the effectiveness of this inverse learning approach.
Stats
MEE = 0.0004
MEE = 0.0001
MEE = 3.1628*10^-5
MEE = 2.4306*10^-5
MEE = 2.4253*10^-5