DuelGAN: A Duel Between Two Discriminators Stabilizes the GAN Training

Abstract

In this paper, we introduce PeerGAN, a generative adversarial network (GAN) solution to improve the stability of the generated samples and to mitigate mode collapse. Built upon the Vanilla GAN's two-player game between the discriminator D 1 and the generator G, we introduce a peer discriminator D 2 to the min-max game. Similar to previous work using two discriminators, the first role of both D 1 , D 2 is to distinguish between generated samples and real ones, while the generator tries to generate high-quality samples which are able to fool both discriminators. Different from existing methods, we introduce another game between D 1 and D 2 to discourage their agreement and therefore increase the level of diversity of the generated samples. This property helps avoid early mode collapse by preventing D 1 and D 2 from converging too fast. We provide theoretical analysis for the equilibrium of the min-max game formed among G, D 1 , D 2 . We offer convergence behavior of PeerGAN as well as stability of the min-max game. It's worth mentioning that PeerGAN operates in the unsupervised setting, and the additional game between D 1 and D 2 does not need any label supervision. Experiments results on a synthetic dataset and on real-world image datasets (MNIST, Fashion MNIST, CIFAR-10, STL-10, CelebA, VGG) demonstrate that PeerGAN outperforms competitive baseline work in generating diverse and high-quality samples, while only introduces negligible computation cost.