Mixed batches and symmetric discriminators for GAN training

TL;DR

This paper introduces a permutation-invariant discriminator architecture and mixed batch training for GANs, effectively reducing mode collapse and improving sample diversity on multiple datasets.

Contribution

It proposes a novel symmetric discriminator architecture and mixed batch training method that enhance GAN training stability and diversity.

Findings

Reduces mode collapse on synthetic datasets

Achieves better qualitative results on CIFAR10 and CelebA

Provides a universal approximation for symmetric functions

Abstract

Generative adversarial networks (GANs) are pow- erful generative models based on providing feed- back to a generative network via a discriminator network. However, the discriminator usually as- sesses individual samples. This prevents the dis- criminator from accessing global distributional statistics of generated samples, and often leads to mode dropping: the generator models only part of the target distribution. We propose to feed the discriminator with mixed batches of true and fake samples, and train it to predict the ratio of true samples in the batch. The latter score does not depend on the order of samples in a batch. Rather than learning this invariance, we introduce a generic permutation-invariant discriminator ar- chitecture. This architecture is provably a uni- versal approximator of all symmetric functions. Experimentally, our approach reduces mode col- lapse in GANs on two synthetic datasets, and obtains good results on the CIFAR10 and CelebA datasets, both qualitatively and quantitatively.