Guardado en:
Detalles Bibliográficos
Autores principales: Huang, Yiwen, Gokaslan, Aaron, Kuleshov, Volodymyr, Tompkin, James
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2501.05441
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913642830626816
author Huang, Yiwen
Gokaslan, Aaron
Kuleshov, Volodymyr
Tompkin, James
author_facet Huang, Yiwen
Gokaslan, Aaron
Kuleshov, Volodymyr
Tompkin, James
contents There is a widely-spread claim that GANs are difficult to train, and GAN architectures in the literature are littered with empirical tricks. We provide evidence against this claim and build a modern GAN baseline in a more principled manner. First, we derive a well-behaved regularized relativistic GAN loss that addresses issues of mode dropping and non-convergence that were previously tackled via a bag of ad-hoc tricks. We analyze our loss mathematically and prove that it admits local convergence guarantees, unlike most existing relativistic losses. Second, our new loss allows us to discard all ad-hoc tricks and replace outdated backbones used in common GANs with modern architectures. Using StyleGAN2 as an example, we present a roadmap of simplification and modernization that results in a new minimalist baseline -- R3GAN. Despite being simple, our approach surpasses StyleGAN2 on FFHQ, ImageNet, CIFAR, and Stacked MNIST datasets, and compares favorably against state-of-the-art GANs and diffusion models.
format Preprint
id arxiv_https___arxiv_org_abs_2501_05441
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The GAN is dead; long live the GAN! A Modern GAN Baseline
Huang, Yiwen
Gokaslan, Aaron
Kuleshov, Volodymyr
Tompkin, James
Machine Learning
Computer Vision and Pattern Recognition
There is a widely-spread claim that GANs are difficult to train, and GAN architectures in the literature are littered with empirical tricks. We provide evidence against this claim and build a modern GAN baseline in a more principled manner. First, we derive a well-behaved regularized relativistic GAN loss that addresses issues of mode dropping and non-convergence that were previously tackled via a bag of ad-hoc tricks. We analyze our loss mathematically and prove that it admits local convergence guarantees, unlike most existing relativistic losses. Second, our new loss allows us to discard all ad-hoc tricks and replace outdated backbones used in common GANs with modern architectures. Using StyleGAN2 as an example, we present a roadmap of simplification and modernization that results in a new minimalist baseline -- R3GAN. Despite being simple, our approach surpasses StyleGAN2 on FFHQ, ImageNet, CIFAR, and Stacked MNIST datasets, and compares favorably against state-of-the-art GANs and diffusion models.
title The GAN is dead; long live the GAN! A Modern GAN Baseline
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2501.05441