Stabilizing Training of Generative Adversarial Networks through Regularization

TL;DR

This paper introduces a new regularization method for GANs that addresses fundamental training stability issues caused by distribution support mismatch, leading to more reliable and stable generative models.

Contribution

The authors propose a novel regularization technique that stabilizes GAN training by overcoming support mismatch problems, with low computational overhead.

Findings

Regularizer improves training stability across various architectures

Enhanced sample quality and convergence consistency

Applicable to common benchmark image generation tasks

Abstract

Deep generative models based on Generative Adversarial Networks (GANs) have demonstrated impressive sample quality but in order to work they require a careful choice of architecture, parameter initialization, and selection of hyper-parameters. This fragility is in part due to a dimensional mismatch or non-overlapping support between the model distribution and the data distribution, causing their density ratio and the associated f-divergence to be undefined. We overcome this fundamental limitation and propose a new regularization approach with low computational cost that yields a stable GAN training procedure. We demonstrate the effectiveness of this regularizer across several architectures trained on common benchmark image generation tasks. Our regularization turns GAN models into reliable building blocks for deep learning.