Nested Annealed Training Scheme for Generative Adversarial Networks

TL;DR

This paper introduces a nested annealed training scheme (NATS) for GANs that enhances sample quality and diversity by integrating a theoretically grounded annealed CFG method, adaptable across various GAN architectures.

Contribution

The paper proposes a novel nested annealed training scheme (NATS) that improves GAN training stability and performance, based on a new theoretical framework linking CFG models and score-based methods.

Findings

Significant improvement in image quality and diversity over baseline GANs.

The nested annealed scheme is adaptable to various GAN architectures.

Experimental results on benchmark datasets validate the effectiveness of the proposed methods.

Abstract

Recently, researchers have proposed many deep generative models, including generative adversarial networks(GANs) and denoising diffusion models. Although significant breakthroughs have been made and empirical success has been achieved with the GAN, its mathematical underpinnings remain relatively unknown. This paper focuses on a rigorous mathematical theoretical framework: the composite-functional-gradient GAN (CFG)[1]. Specifically, we reveal the theoretical connection between the CFG model and score-based models. We find that the training objective of the CFG discriminator is equivalent to finding an optimal D(x). The optimal gradient of D(x) differentiates the integral of the differences between the score functions of real and synthesized samples. Conversely, training the CFG generator involves finding an optimal G(x) that minimizes this difference. In this paper, we aim to derive an annealed weight preceding the weight of the CFG discriminator. This new explicit theoretical explanation model is called the annealed CFG method. To overcome the limitation of the annealed CFG method, as the method is not readily applicable to the SOTA GAN model, we propose a nested annealed training scheme (NATS). This scheme keeps the annealed weight from the CFG method and can be seamlessly adapted to various GAN models, no matter their structural, loss, or regularization differences. We conduct thorough experimental evaluations on various benchmark datasets for image generation. The results show that our annealed CFG and NATS methods significantly improve the quality and diversity of the synthesized samples. This improvement is clear when comparing the CFG method and the SOTA GAN models.