A potentialization algorithm for games with applications to multi-agent learning in repeated games

TL;DR

This paper introduces a potentialization algorithm that transforms any normal form game into a potential game, enabling efficient multi-agent learning and convergence to stable behaviors through surrogate rewards.

Contribution

The paper presents a novel algorithm that constructs potential games from arbitrary normal form games, facilitating improved learning dynamics in multi-agent systems.

Findings

Potentialization guarantees convergence to stable agent behavior.

Numerical simulations confirm effectiveness with replicator dynamics.

Algorithm broadens applicability of potential games in multi-agent learning.

Abstract

We investigate an algorithm that assigns to any game in normal form an approximating game that admits an ordinal potential function. Due to the properties of potential games, the algorithm equips every game with a surrogate reward structure that allows efficient multi-agent learning. Numerical simulations using the replicator dynamics show that 'potentialization' guarantees convergence to stable agent behavior.