Improving the Improved Training of Wasserstein GANs: A Consistency Term and Its Dual Effect

TL;DR

This paper introduces a new method for enforcing Lipschitz continuity in Wasserstein GANs, leading to improved image quality and state-of-the-art semi-supervised learning results with fewer labeled data.

Contribution

It proposes a novel approach to enforce Lipschitz continuity in WGANs, connecting GAN training with semi-supervised learning, and achieves superior results.

Findings

Higher inception scores on CIFAR-10 with fewer images

Achieves over 90% accuracy with only 4,000 labeled images

Produces more photo-realistic samples than previous methods

Abstract

Despite being impactful on a variety of problems and applications, the generative adversarial nets (GANs) are remarkably difficult to train. This issue is formally analyzed by \cite{arjovsky2017towards}, who also propose an alternative direction to avoid the caveats in the minmax two-player training of GANs. The corresponding algorithm, called Wasserstein GAN (WGAN), hinges on the 1-Lipschitz continuity of the discriminator. In this paper, we propose a novel approach to enforcing the Lipschitz continuity in the training procedure of WGANs. Our approach seamlessly connects WGAN with one of the recent semi-supervised learning methods. As a result, it gives rise to not only better photo-realistic samples than the previous methods but also state-of-the-art semi-supervised learning results. In particular, our approach gives rise to the inception score of more than 5.0 with only 1,000 CIFAR-10 images and is the first that exceeds the accuracy of 90% on the CIFAR-10 dataset using only 4,000 labeled images, to the best of our knowledge.