改进的Wasserstein GANs训练

Mar, 2017

Improved Training of Wasserstein GANs

Ishaan Gulrajani, Faruk Ahmed, Martin Arjovsky, Vincent Dumoulin, Aaron Courville

TL;DR提出了一种比传统的WGAN更好的GAN训练方法，使用正则化替代权重截取，通过惩罚评判器对其输入梯度的范数，可以实现各种GAN结构的稳定训练和高质量生成。

Abstract

generative adversarial networks (GANs) are powerful generative models, but suffer from training instability. The recently proposed wasserstein gan (WGAN) makes significant progress toward stable training of GANs,