GANs May Have No Nash Equilibria

Abstract

Generative adversarial networks (GANs) represent a zero-sum game between two machine players, a generator and a discriminator, designed to learn the distribution of data. While GANs have achieved state-of-the-art performance in several benchmark learning tasks, GAN minimax optimization still poses great theoretical and empirical challenges. GANs trained using first-order optimization methods commonly fail to converge to a stable solution where the players cannot improve their objective, i.e., the Nash equilibrium of the underlying game. Such issues raise the question of the existence of Nash equilibrium solutions in the GAN zero-sum game. In this work, we show through several theoretical and numerical results that indeed GAN zero-sum games may not have any local Nash equilibria. To characterize an equilibrium notion applicable to GANs, we consider the equilibrium of a new zero-sum game with an objective function given by a proximal operator applied to the original objective, a solution we call the proximal equilibrium. Unlike the Nash equilibrium, the proximal equilibrium captures the sequential nature of GANs, in which the generator moves first followed by the discriminator. We prove that the optimal generative model in Wasserstein GAN problems provides a proximal equilibrium. Inspired by these results, we propose a new approach, which we call proximal training, for solving GAN problems. We discuss several numerical experiments demonstrating the existence of proximal equilibrium solutions in GAN minimax problems.

References in corpus (24)

• Adam: A Method for Stochastic Optimization
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Spectral Normalization for Generative Adversarial Networks
• NIPS 2016 Tutorial: Generative Adversarial Networks
• BEGAN: Boundary Equilibrium Generative Adversarial Networks
• On Convergence and Stability of GANs
• Generalization and Equilibrium in Generative Adversarial Nets (GANs)
• On Gradient Descent Ascent for Nonconvex-Concave Minimax Problems
• On the Convergence and Robustness of Adversarial Training
• The Numerics of GANs
• Solving a Class of Non-Convex Min-Max Games Using Iterative First Order Methods
• Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Step
• On Finding Local Nash Equilibria (and Only Local Nash Equilibria) in Zero-Sum Games
• Policy Optimization Provably Converges to Nash Equilibria in Zero-Sum Linear Quadratic Games
• What is Local Optimality in Nonconvex-Nonconcave Minimax Optimization?
• On the Convergence and Robustness of Training GANs with Regularized Optimal Transport
• Improving GANs Using Optimal Transport
• Convergence of Learning Dynamics in Stackelberg Games
• Lipschitz Generative Adversarial Nets
• Understanding GANs: the LQG Setting
• 2-Wasserstein Approximation via Restricted Convex Potentials with Application to Improved Training for GANs
• Finding Mixed Nash Equilibria of Generative Adversarial Networks
• SGD Learns One-Layer Networks in WGANs

Cited by in corpus (11)

• Investigating Cosmological GAN Emulators Using Latent Space Interpolation
• Implicit competitive regularization in GANs
• Understanding Overparameterization in Generative Adversarial Networks
• On the Suboptimality of Negative Momentum for Minimax Optimization
• A Decentralized Adaptive Momentum Method for Solving a Class of Min-Max Optimization Problems
• Train simultaneously, generalize better: Stability of gradient-based minimax learners
• Nonconvex-Nonconcave Min-Max Optimization with a Small Maximization Domain
• Multi-Modal Super Resolution for Dense Microscopic Particle Size Estimation
• Game of GANs: Game-Theoretical Models for Generative Adversarial Networks
• Testing and Emulating Modified Gravity on Cosmological Scales
• Solving Min-Max Optimization with Hidden Structure via Gradient Descent Ascent