Training GANs with Optimism

TL;DR

This paper introduces Optimistic Mirror Descent (OMD) for training Wasserstein GANs, addressing limit cycling issues, and demonstrates improved convergence and performance over traditional methods in both theoretical and practical settings.

Contribution

It proposes OMD for WGAN training, proves its convergence in zero-sum games, and introduces Optimistic Adam, showing empirical improvements over existing algorithms.

Findings

OMD addresses limit cycling in WGAN training.

Models trained with OMD have smaller KL divergence.

Optimistic Adam improves CIFAR10 performance.

Abstract

We address the issue of limit cycling behavior in training Generative Adversarial Networks and propose the use of Optimistic Mirror Decent (OMD) for training Wasserstein GANs. Recent theoretical results have shown that optimistic mirror decent (OMD) can enjoy faster regret rates in the context of zero-sum games. WGANs is exactly a context of solving a zero-sum game with simultaneous no-regret dynamics. Moreover, we show that optimistic mirror decent addresses the limit cycling problem in training WGANs. We formally show that in the case of bi-linear zero-sum games the last iterate of OMD dynamics converges to an equilibrium, in contrast to GD dynamics which are bound to cycle. We also portray the huge qualitative difference between GD and OMD dynamics with toy examples, even when GD is modified with many adaptations proposed in the recent literature, such as gradient penalty or momentum. We apply OMD WGAN training to a bioinformatics problem of generating DNA sequences. We observe that models trained with OMD achieve consistently smaller KL divergence with respect to the true underlying distribution, than models trained with GD variants. Finally, we introduce a new algorithm, Optimistic Adam, which is an optimistic variant of Adam. We apply it to WGAN training on CIFAR10 and observe improved performance in terms of inception score as compared to Adam.