Software Agents for Learning Nash Equilibria in Non-Cooperative Games

This paper describes SALENE, a Multi Agent System (MAS) for learning Nash Equilibria in non-cooperative games. SALENE is based on the following assumptions: if agents representing the players act as rational players, i.e. they act to maximise their expected utility in each match of a game, and if such agents play k matches of the game they will converge in playing one of the Nash Equilibria of the game. SALENE can be conceived as a heuristic and efficient method to compute at least one Nash Equilibria in a non-cooperative game represented in its normal form.