Potential Games
Overview
Potential games represent how an individual player's change in strategy relates to a shared scalar function, making stable outcomes easier to analyze and compute. Methods include exact, ordinal, weighted, and α-potential constructions, alongside best-response dynamics, policy-gradient learning, zeroth-order optimization, and distributed primal–dual algorithms. Applications range from federated learning and multi-agent reinforcement learning to traffic coordination, resource allocation, coalition formation, and scheduling. A central challenge is that approximating asymmetric general-sum games with a common potential can ease equilibrium search but distort incentives, equilibrium accuracy, or social efficiency.
Latest papers
17 papers1 paper in the last four weeks, down 67% on the four weeks before. 0.0% of all new papers.