Some theoretical properties of GANs

DOI10.1214/19-AOS1858zbMATH Open1451.62107arXiv1803.07819OpenAlexW3043551088MaRDI QIDQ2196234FDOQ2196234

Authors: Gérard Biau, Benoît Cadre, Maxime Sangnier, Ugo Tanielian 

Publication date: 28 August 2020

Published in: The Annals of Statistics (Search for Journal in Brave)

Abstract: Generative Adversarial Networks (GANs) are a class of generative algorithms that have been shown to produce state-of-the art samples, especially in the domain of image creation. The fundamental principle of GANs is to approximate the unknown distribution of a given data set by optimizing an objective function through an adversarial game between a family of generators and a family of discriminators. In this paper, we offer a better theoretical understanding of GANs by analyzing some of their mathematical and statistical properties. We study the deep connection between the adversarial principle underlying GANs and the Jensen-Shannon divergence, together with some optimality characteristics of the problem. An analysis of the role of the discriminator family via approximation arguments is also provided. In addition, taking a statistical point of view, we study the large sample properties of the estimated distribution and prove in particular a central limit theorem. Some of our results are illustrated with simulated examples.

Full work available at URL: https://arxiv.org/abs/1803.07819

Recommendations

• Some theoretical insights into Wasserstein GANs
• Deconstructing Generative Adversarial Networks
• Understanding Estimation and Generalization Error of Generative Adversarial Networks
• scientific article; zbMATH DE number 7267187
• Wasserstein proximal of GANs
• Generative modeling of convolutional neural networks
• Generalization error of GAN from the discriminator's perspective
• scientific article; zbMATH DE number 7626743

zbMATH Keywords

central limit theoremJensen-Shannon divergenceneural networksgenerative modelsadversarial principle

Mathematics Subject Classification ID

Asymptotic properties of parametric estimators (62F12) Central limit and other weak theorems (60F05) Neural nets and related approaches to inference from stochastic processes (62M45) 2-person games (91A05) General topics in artificial intelligence (68T01)

Cites Work

• Title not available (Why is that?)
• A new metric for probability distributions
• Universal smoothing factor selection in density estimation: theory and practice. (With discussion)

Cited In (22)

• GANs for generating EFT models
• Least \(k\)th-order and Rényi generative adversarial networks
• Multi-projection of unequal dimension optimal transport theory for generative adversary networks
• Generative adversarial networks with joint distribution moment matching
• Posterior asymptotics in Wasserstein metrics on the real line
• A convenient infinite dimensional framework for generative adversarial learning
• Training GANs with centripetal acceleration
• Optimal 1-Wasserstein distance for WGANs
• Statistical inference for generative adversarial networks and other minimax problems
• Title not available (Why is that?)
• Adversarial perturbations of deep neural networks
• Monte Carlo simulation of SDEs using GANs
• A survey of image synthesis and editing with generative adversarial networks
• Discriminator feature-based inference by recycling the discriminator of GANs
• Minimax rate of distribution estimation on unknown submanifolds under adversarial losses
• A comprehensive survey and analysis of generative models in machine learning
• Stochastic differential equation approximations of generative adversarial network training and its long-run behavior
• Generalization error of GAN from the discriminator's perspective
• Finite sample properties of parametric MMD estimation: robustness to misspecification and dependence
• Geometrical Insights for Implicit Generative Modeling
• Enforcing imprecise constraints on generative adversarial networks for emulating physical systems
• GAT–GMM: Generative Adversarial Training for Gaussian Mixture Models

Uses Software

• TensorFlow
• Wasserstein GAN
• VEEGAN
• f-GAN

This page was built for publication: Some theoretical properties of GANs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2196234)