Research Article
Learning a Correlated Equilibrium with Perturbed Regret Minimization
@INPROCEEDINGS{10.1007/978-3-031-31234-2_2, author={Omar Boufous and Rachid El-Azouzi and Mika\`{\i}l Touati and Eitan Altman and Mustapha Bouhtou}, title={Learning a Correlated Equilibrium with Perturbed Regret Minimization}, proceedings={Performance Evaluation Methodologies and Tools. 15th EAI International Conference, VALUETOOLS 2022, Virtual Event, November 2022, Proceedings}, proceedings_a={VALUETOOLS}, year={2023}, month={5}, keywords={Game theory Correlated equilibrium Learning}, doi={10.1007/978-3-031-31234-2_2} }- Omar Boufous
Rachid El-Azouzi
Mikaël Touati
Eitan Altman
Mustapha Bouhtou
Year: 2023
Learning a Correlated Equilibrium with Perturbed Regret Minimization
VALUETOOLS
Springer
DOI: 10.1007/978-3-031-31234-2_2
Abstract
In this paper, we consider the problem of learning a correlated equilibrium of a finite non-cooperative game and show a new learning rule, called Correlated Perturbed Regret Minimization (CPRM) for this purpose. CPRM combines regret minimization to approach the set of correlated equilibria and a simple device recommending to the players actions drawn from the empirical distribution in order to further stabilize the dynamic. Numerical experiments support the hypothesis of the pointwise convergence of the empirical distribution over action profiles to an approximate correlated equilibrium with all players following the devices’ suggestions. Additional simulation results suggest that an adaptive version of CPRM can handle changes in the game such as departures or arrivals of players.
