UCD: Unconditional Discriminator Promotes Nash Equilibrium in GANs
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914069993226240 |
|---|---|
| author | Xia, Mengfei Xue, Nan Zhu, Jiapeng Shen, Yujun |
| author_facet | Xia, Mengfei Xue, Nan Zhu, Jiapeng Shen, Yujun |
| contents | Adversarial training turns out to be the key to one-step generation, especially for Generative Adversarial Network (GAN) and diffusion model distillation. Yet in practice, GAN training hardly converges properly and struggles in mode collapse. In this work, we quantitatively analyze the extent of Nash equilibrium in GAN training, and conclude that redundant shortcuts by inputting condition in $D$ disables meaningful knowledge extraction. We thereby propose to employ an unconditional discriminator (UCD), in which $D$ is enforced to extract more comprehensive and robust features with no condition injection. In this way, $D$ is able to leverage better knowledge to supervise $G$, which promotes Nash equilibrium in GAN literature. Theoretical guarantee on compatibility with vanilla GAN theory indicates that UCD can be implemented in a plug-in manner. Extensive experiments confirm the significant performance improvements with high efficiency. For instance, we achieved \textbf{1.47 FID} on the ImageNet-64 dataset, surpassing StyleGAN-XL and several state-of-the-art one-step diffusion models. The code will be made publicly available. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_00624 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | UCD: Unconditional Discriminator Promotes Nash Equilibrium in GANs Xia, Mengfei Xue, Nan Zhu, Jiapeng Shen, Yujun Computer Vision and Pattern Recognition Adversarial training turns out to be the key to one-step generation, especially for Generative Adversarial Network (GAN) and diffusion model distillation. Yet in practice, GAN training hardly converges properly and struggles in mode collapse. In this work, we quantitatively analyze the extent of Nash equilibrium in GAN training, and conclude that redundant shortcuts by inputting condition in $D$ disables meaningful knowledge extraction. We thereby propose to employ an unconditional discriminator (UCD), in which $D$ is enforced to extract more comprehensive and robust features with no condition injection. In this way, $D$ is able to leverage better knowledge to supervise $G$, which promotes Nash equilibrium in GAN literature. Theoretical guarantee on compatibility with vanilla GAN theory indicates that UCD can be implemented in a plug-in manner. Extensive experiments confirm the significant performance improvements with high efficiency. For instance, we achieved \textbf{1.47 FID} on the ImageNet-64 dataset, surpassing StyleGAN-XL and several state-of-the-art one-step diffusion models. The code will be made publicly available. |
| title | UCD: Unconditional Discriminator Promotes Nash Equilibrium in GANs |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2510.00624 |