Decentralized Fictitious Play in Near-Potential Games With Time-Varying Communication Networks

We study the convergence properties of decentralized fictitious play (DFP) for class near-potential games where incentives agents are nearly aligned with a potential function. In DFP, share information only their current neighbors in sequence time-varying networks, keep estimates other agents' empirical frequencies, and take actions to maximize expected utility functions computed respect estimated frequencies. show that frequencies converge set strategies function values larger than obtained by approximate Nash equilibria game. This result establishes DFP has identical guarantees as standard which observe past all agents.