Turbocharging Solution Concepts: Solving NEs, CEs and CCEs with Neural Equilibrium Solvers

TL;DR

This paper introduces a neural network-based solver for efficiently approximating Nash, Correlated, and Coarse Correlated Equilibria in multiagent games, enabling fast, deterministic solutions with strong generalization.

Contribution

The authors develop a neural equilibrium solver with equivariant architecture that can solve all fixed-shape games quickly, without supervised data, and generalizes to larger games.

Findings

Achieves zero-shot generalization to larger games.

Provides a flexible equilibrium selection framework.

Offers a fast, deterministic alternative to traditional methods.

Abstract

Solution concepts such as Nash Equilibria, Correlated Equilibria, and Coarse Correlated Equilibria are useful components for many multiagent machine learning algorithms. Unfortunately, solving a normal-form game could take prohibitive or non-deterministic time to converge, and could fail. We introduce the Neural Equilibrium Solver which utilizes a special equivariant neural network architecture to approximately solve the space of all games of fixed shape, buying speed and determinism. We define a flexible equilibrium selection framework, that is capable of uniquely selecting an equilibrium that minimizes relative entropy, or maximizes welfare. The network is trained without needing to generate any supervised training data. We show remarkable zero-shot generalization to larger games. We argue that such a network is a powerful component for many possible multiagent algorithms.