Improving GAN Training via Feature Space Shrinkage

TL;DR

This paper introduces AdaptiveMix, a module that improves GAN training stability and image quality by shrinking feature space regions, and demonstrates its effectiveness across multiple tasks and datasets.

Contribution

Proposes AdaptiveMix, a novel module that constrains feature space in GANs, enhancing training stability and sample quality, and extends its application to classification and OOD detection.

Findings

AdaptiveMix improves GAN image quality across architectures.

AdaptiveMix enhances robustness in image classification.

AdaptiveMix boosts OOD detection performance.

Abstract

Due to the outstanding capability for data generation, Generative Adversarial Networks (GANs) have attracted considerable attention in unsupervised learning. However, training GANs is difficult, since the training distribution is dynamic for the discriminator, leading to unstable image representation. In this paper, we address the problem of training GANs from a novel perspective, \emph{i.e.,} robust image classification. Motivated by studies on robust image representation, we propose a simple yet effective module, namely AdaptiveMix, for GANs, which shrinks the regions of training data in the image representation space of the discriminator. Considering it is intractable to directly bound feature space, we propose to construct hard samples and narrow down the feature distance between hard and easy samples. The hard samples are constructed by mixing a pair of training images. We evaluate the effectiveness of our AdaptiveMix with widely-used and state-of-the-art GAN architectures. The evaluation results demonstrate that our AdaptiveMix can facilitate the training of GANs and effectively improve the image quality of generated samples. We also show that our AdaptiveMix can be further applied to image classification and Out-Of-Distribution (OOD) detection tasks, by equipping it with state-of-the-art methods. Extensive experiments on seven publicly available datasets show that our method effectively boosts the performance of baselines. The code is publicly available at https://github.com/WentianZhang-ML/AdaptiveMix.