Adaptive adversarial training method for improving multi-scale GAN based on generalization bound theory

TL;DR

This paper introduces an adaptive adversarial training method for multi-scale GANs, grounded in PAC-Bayes theory, significantly enhancing image quality and manipulation capabilities, especially in super-resolution tasks.

Contribution

The paper pioneers the integration of PAC-Bayes generalized bound theory into multi-scale GAN training, proposing an adaptive method that improves image generation quality and generalization.

Findings

Significant reduction in NIQE and RMSE for super-resolution tasks

Improved image manipulation ability of multi-scale GANs

Enhanced generalization error bounds under various attacks

Abstract

In recent years, multi-scale generative adversarial networks (GANs) have been proposed to build generalized image processing models based on single sample. Constraining on the sample size, multi-scale GANs have much difficulty converging to the global optimum, which ultimately leads to limitations in their capabilities. In this paper, we pioneered the introduction of PAC-Bayes generalized bound theory into the training analysis of specific models under different adversarial training methods, which can obtain a non-vacuous upper bound on the generalization error for the specified multi-scale GAN structure. Based on the drastic changes we found of the generalization error bound under different adversarial attacks and different training states, we proposed an adaptive training method which can greatly improve the image manipulation ability of multi-scale GANs. The final experimental results show that our adaptive training method in this paper has greatly contributed to the improvement of the quality of the images generated by multi-scale GANs on several image manipulation tasks. In particular, for the image super-resolution restoration task, the multi-scale GAN model trained by the proposed method achieves a 100% reduction in natural image quality evaluator (NIQE) and a 60% reduction in root mean squared error (RMSE), which is better than many models trained on large-scale datasets.