arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2008.09041eess.IVcs.CVcs.LG
收藏

稳定GANs训练的新视角:直接对抗训练

A New Perspective on Stabilizing GANs training: Direct Adversarial Training

University of Science and Technology of China(中国科学技术大学)

Ziqiang Li, Pengfei Xia, Rentuo Tao, Hongjing Niu, Bin Li

摘要:

生成对抗网络(GANs)是最流行的图像生成模型,在各种计算机视觉任务上取得了显着进展。然而,训练不稳定性仍然是所有基于GAN算法的开放问题之一。目前已经提出了许多方法来稳定GANs的训练,其重点分别放在损失函数,正则化和归一化技术,训练算法和模型结构。与上述方法不同,本文提出了一种新的稳定GANs训练的视角。研究发现,有时生成器产生的图像在训练过程中表现得像是训练器的对抗样本,这可能是导致GANs训练不稳定的部分原因。有了这一发现,我们提出了直接对抗训练(DAT)方法来稳定GANs的训练过程。此外,我们证明了DAT方法是能够最小化的Lipschitz常数的自适应。DAT的先进性能在多个损失函数、网络架构、超参数和数据集上得到了验证。具体而言,DAT基于SSGAN在CIFAR-100无条件生成上实现了11.5%的FID,基于SSGAN在STL-10无条件生成上实现了10.5%的FID,基于SSGAN在LSUN-Bedroom无条件生成上实现了13.2%的FID。代码将在https://github.com/iceli1007/DAT-GAN上提供

英文摘要:

Generative Adversarial Networks (GANs) are the most popular image generation models that have achieved remarkable progress on various computer vision tasks. However, training instability is still one of the open problems for all GAN-based algorithms. Quite a number of methods have been proposed to stabilize the training of GANs, the focuses of which were respectively put on the loss functions, regularization and normalization technologies, training algorithms, and model architectures. Different from the above methods, in this paper, a new perspective on stabilizing GANs training is presented. It is found that sometimes the images produced by the generator act like adversarial examples of the discriminator during the training process, which may be part of the reason causing the unstable training of GANs. With this finding, we propose the Direct Adversarial Training (DAT) method to stabilize the training process of GANs. Furthermore, we prove that the DAT method is able to minimize the Lipschitz constant of the discriminator adaptively. The advanced performance of DAT is verified on multiple loss functions, network architectures, hyper-parameters, and datasets. Specifically, DAT achieves significant improvements of 11.5% FID on CIFAR-100 unconditional generation based on SSGAN, 10.5% FID on STL-10 unconditional generation based on SSGAN, and 13.2% FID on LSUN-Bedroom unconditional generation based on SSGAN. Code will be available at https://github.com/iceli1007/DAT-GAN

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑