Understanding Dynamics of Adam in Zero-Sum Games: An ODE Approach

TL;DR

This paper derives ODEs as continuous limits of Adam-DA in zero-sum games, providing insights into its convergence and regularization effects, and validates findings with GAN experiments.

Contribution

It introduces an ODE framework for Adam-DA in zero-sum games, revealing reversed roles of momentum parameters and offering new theoretical insights.

Findings

ODEs closely approximate Adam-DA dynamics

Reversed effects of momentum parameters in zero-sum games

Validated predictions through GAN experiments

Abstract

The remarkable success of the Adam in training neural networks has naturally led to the widespread use of its descent-ascent counterpart, Adam-DA, for solving zero-sum games. Despite its popularity in practice, a rigorous theoretical understanding of Adam-DA still lags behind. In this paper, we derive ordinary differential equations (ODEs) that serve as continuous-time limits of the Adam-DA. These ODEs closely approximate the discrete-time dynamics of Adam-DA, providing a tractable analytical framework for understanding its behavior in zero-sum games. Using this ODE approach, we investigate two fundamental aspects of Adam-DA: local convergence and implicit gradient regularization. Our analysis reveals that the roles of the first- and second-order momentum parameters in zero-sum games are exactly the opposite of their well-documented effects in minimization problems. We validate these predictions through GAN experiments across multiple architectures and datasets, demonstrating the practical implications of this reversed momentum effect.