This paper introduces a novel decentralized learning approach for finite normal-form games over dynamic networks. The method allows agents to learn socially optimal equilibria without prior knowledge of the game, relying only on local payoff comparisons and communication with time-varying neighbors. Agents exchange randomized signals and time-stamped tables, utilizing table fusion and temporal majority reconstruction to handle dynamic communication while maintaining decentralized operation. The approach achieves finite-time logarithmic regret guarantees for optimal equilibrium selection under utilitarian and proportional-fair social welfare objectives, with simulations demonstrating its effectiveness. AI
IMPACT Introduces a new method for decentralized learning in game theory, potentially applicable to multi-agent systems and distributed AI.
RANK_REASON The cluster contains a research paper published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
