Abstract

Recent work has shown local convergence of GAN training for absolutely continuous data and generator distributions. In this paper, we show that the requirement of absolute continuity is nec essary: we describe a simple yet prototypical counterexample showing that in the more realistic case of distributions that are not absolutel continuous, unregularized GAN training is not always convergent. Furthermore, we discuss regularization strategies that were recently proposed to stabilize GAN training. Our analysis shows that GAN training with instance noise or zerocentered gradient penalties converges. On the other hand, we show that Wasserstein-GANs and WGAN-GP with a finite number of discriminator updates per generator update do not always converge to the equilibrium point. We discuss these results, leading us to a new explanation for the stability problems of GAN training. Based on our analysis, we extend our convergence results to more general GANs and prove local convergence for simplified gradient penalties even if th generator and data distributions lie on lower dimensional manifolds. We find these penalties to work well in practice and use them to learn highresolution generative image models for a variety of datasets with little hyperparameter tuning.