A Neural Tangent Kernel Perspective of GANs - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2022

A Neural Tangent Kernel Perspective of GANs

Résumé

We propose a novel theoretical framework of analysis for Generative Adversarial Networks (GANs). We start by pointing out a fundamental flaw in previous theoretical analyses that leads to ill-defined gradients for the discriminator. We overcome this issue which impedes a principled study of GAN training, solving it within our framework by taking into account the discriminator's architecture. To this end, we leverage the theory of infinite-width neural networks for the discriminator via its Neural Tangent Kernel. We characterize the trained discriminator for a wide range of losses and establish general differentiability properties of the network. Moreover, we derive new insights about the generated distribution's flow during training, advancing our understanding of GAN dynamics. We empirically corroborate these results via a publicly released analysis toolkit based on our framework, unveiling intuitions that are consistent with current GAN practice.
Fichier principal
Vignette du fichier
gantk2.pdf (2.91 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03254591 , version 1 (08-06-2021)
hal-03254591 , version 2 (20-10-2021)
hal-03254591 , version 3 (31-01-2022)
hal-03254591 , version 4 (15-06-2022)
hal-03254591 , version 5 (27-10-2022)

Licence

Paternité

Identifiants

Citer

Jean-Yves Franceschi, Emmanuel de Bézenac, Ibrahim Ayed, Mickaël Chen, Sylvain Lamprier, et al.. A Neural Tangent Kernel Perspective of GANs. 2022. ⟨hal-03254591v3⟩
265 Consultations
237 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More