Stabilizing Training of Generative Adversarial Nets via Langevin Stein Variational Gradient Descent

TL;DR

This paper introduces Langevin Stein Variational Gradient Descent (LSVGD), a novel particle-based method to stabilize GAN training by enhancing diversity and convergence, leading to improved performance on multiple datasets.

Contribution

The paper proposes LSVGD, a new particle-based variational inference method that stabilizes GAN training by incorporating noise and regularization, addressing instability issues of existing approaches.

Findings

LSVGD improves GAN stability and performance.

Experimental results on multiple datasets show enhanced diversity.

LSVGD effectively integrates with various GAN loss functions.

Abstract

Generative adversarial networks (GANs), famous for the capability of learning complex underlying data distribution, are however known to be tricky in the training process, which would probably result in mode collapse or performance deterioration. Current approaches of dealing with GANs' issues almost utilize some practical training techniques for the purpose of regularization, which on the other hand undermines the convergence and theoretical soundness of GAN. In this paper, we propose to stabilize GAN training via a novel particle-based variational inference -- Langevin Stein variational gradient descent (LSVGD), which not only inherits the flexibility and efficiency of original SVGD but aims to address its instability issues by incorporating an extra disturbance into the update dynamics. We further demonstrate that by properly adjusting the noise variance, LSVGD simulates a Langevin process whose stationary distribution is exactly the target distribution. We also show that LSVGD dynamics has an implicit regularization which is able to enhance particles' spread-out and diversity. At last we present an efficient way of applying particle-based variational inference on a general GAN training procedure no matter what loss function is adopted. Experimental results on one synthetic dataset and three popular benchmark datasets -- Cifar-10, Tiny-ImageNet and CelebA validate that LSVGD can remarkably improve the performance and stability of various GAN models.