Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Step

Abstract

Generative adversarial networks (GANs) are a family of generative models that do not minimize a single training criterion. Unlike other generative models, the data distribution is learned via a game between a generator (the generative model) and a discriminator (a teacher providing training signal) that each minimize their own cost. GANs are designed to reach a Nash equilibrium at which each player cannot reduce their cost without changing the other players' parameters. One useful approach for the theory of GANs is to show that a divergence between the training distribution and the model distribution obtains its minimum value at equilibrium. Several recent research directions have been motivated by the idea that this divergence is the primary guide for the learning process and that every step of learning should decrease the divergence. We show that this view is overly restrictive. During GAN training, the discriminator provides learning signal in situations where the gradients of the divergences between distributions would not be useful. We provide empirical counterexamples to the view of GAN training as divergence minimization. Specifically, we demonstrate that GANs are able to learn distributions in situations where the divergence minimization point of view predicts they would fail. We also show that gradient penalties motivated from the divergence minimization perspective are equally helpful when applied in other contexts in which the divergence minimization perspective does not predict they would be helpful. This contributes to a growing body of evidence that GAN training may be more usefully viewed as approaching Nash equilibria via trajectories that do not necessarily minimize a specific divergence at each step.

18 pages

References in corpus (5)

• Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks
• Towards Principled Methods for Training Generative Adversarial Networks
• Variational Approaches for Auto-Encoding Generative Adversarial Networks
• Amortised MAP Inference for Image Super-resolution
• Comparison of Maximum Likelihood and GAN-based training of Real NVPs

Cited by in corpus (35)

• The relativistic discriminator: a key element missing from standard GAN
• Quantum Generative Adversarial Networks for Learning and Loading Random Distributions
• A DIRT-T Approach to Unsupervised Domain Adaptation
• Unsupervised Domain Adaptation through Self-Supervision
• Enhanced Balancing GAN: Minority-class Image Generation
• Diagnosing and Enhancing VAE Models
• A Systematic Survey of Regularization and Normalization in GANs
• Convergence Problems with Generative Adversarial Networks (GANs)
• Improving Detection of Credit Card Fraudulent Transactions using Generative Adversarial Networks
• MGGAN: Solving Mode Collapse using Manifold Guided Training
• Improved Training of Generative Adversarial Networks Using Representative Features
• GANs May Have No Nash Equilibria
• Sample-Efficient Imitation Learning via Generative Adversarial Nets
• Exploring the Evolution of GANs through Quality Diversity
• Age-Oriented Face Synthesis with Conditional Discriminator Pool and Adversarial Triplet Loss
• Noise Adaptive Speech Enhancement using Domain Adversarial Training
• Near-Term Quantum-Classical Associative Adversarial Networks
• ChildGAN: Large Scale Synthetic Child Facial Data Using Domain Adaptation in StyleGAN
• Primal-Dual Wasserstein GAN
• Understanding the Effectiveness of Lipschitz-Continuity in Generative Adversarial Nets
• Generative Adversarial Zero-Shot Relational Learning for Knowledge Graphs
• GANs beyond divergence minimization
• First Order Generative Adversarial Networks
• Cross-Domain Sentiment Classification with In-Domain Contrastive Learning
• Resembled Generative Adversarial Networks: Two Domains with Similar Attributes
• Decomposed Adversarial Learned Inference
• Collaborative Sampling in Generative Adversarial Networks
• 3D Human motion anticipation and classification
• A Novel Framework for Selection of GANs for an Application
• Game of GANs: Game-Theoretical Models for Generative Adversarial Networks
• Online Kernel based Generative Adversarial Networks
• Cross-Domain Sentiment Classification with Contrastive Learning and Mutual Information Maximization
• Exploiting the Hidden Tasks of GANs: Making Implicit Subproblems Explicit
• MMCGAN: Generative Adversarial Network with Explicit Manifold Prior
• Wavelets to the Rescue: Improving Sample Quality of Latent Variable Deep Generative Models