A Neural Tangent Kernel Perspective of GANs

06/10/2021
by   Jean-Yves Franceschi, et al.
8

Theoretical analyses for Generative Adversarial Networks (GANs) generally assume an arbitrarily large family of discriminators and do not consider the characteristics of the architectures used in practice. We show that this framework of analysis is too simplistic to properly analyze GAN training. To tackle this issue, we leverage the theory of infinite-width neural networks to model neural discriminator training for a wide range of adversarial losses via its Neural Tangent Kernel (NTK). Our analytical results show that GAN trainability primarily depends on the discriminator's architecture. We further study the discriminator for specific architectures and losses, and highlight properties providing a new understanding of GAN training. For example, we find that GANs trained with the integral probability metric loss minimize the maximum mean discrepancy with the NTK as kernel. Our conclusions demonstrate the analysis opportunities provided by the proposed framework, which paves the way for better and more principled GAN models. We release a generic GAN analysis toolkit based on our framework that supports the empirical part of our study.

READ FULL TEXT

page 8

page 9

page 34

page 35

research
06/02/2023

GANs Settle Scores!

Generative adversarial networks (GANs) comprise a generator, trained to ...
research
12/18/2019

Lower Dimensional Kernels for Video Discriminators

This work presents an analysis of the discriminators used in Generative ...
research
01/04/2018

Demystifying MMD GANs

We investigate the training and performance of generative adversarial ne...
research
11/27/2020

A study of traits that affect learnability in GANs

Generative Adversarial Networks GANs are algorithmic architectures that ...
research
10/03/2020

Assisting the Adversary to Improve GAN Training

We propose a method for improved training of generative adversarial netw...
research
07/12/2018

The GAN Landscape: Losses, Architectures, Regularization, and Normalization

Generative Adversarial Networks (GANs) are a class of deep generative mo...
research
02/15/2018

Selecting the Best in GANs Family: a Post Selection Inference Framework

"Which Generative Adversarial Networks (GANs) generates the most plausib...

Please sign up or login with your details

Forgot password? Click here to reset