Perturbative GAN: GAN with Perturbation Layers

TL;DR

Perturbative GAN introduces fixed noise perturbation layers in place of convolutional layers, reducing parameters, accelerating training, and improving image quality, with efficient noise mask generation for hardware-accelerated training.

Contribution

This paper proposes a novel GAN architecture replacing convolution layers with perturbation layers, enhancing training speed, reducing parameters, and improving image quality.

Findings

Faster training convergence compared to traditional GANs

Higher inception scores on standard datasets

Reduced training cost and parameter count

Abstract

Perturbative GAN, which replaces convolution layers of existing convolutional GANs (DCGAN, WGAN-GP, BIGGAN, etc.) with perturbation layers that adds a fixed noise mask, is proposed. Compared with the convolu-tional GANs, the number of parameters to be trained is smaller, the convergence of training is faster, the incep-tion score of generated images is higher, and the overall training cost is reduced. Algorithmic generation of the noise masks is also proposed, with which the training, as well as the generation, can be boosted with hardware acceleration. Perturbative GAN is evaluated using con-ventional datasets (CIFAR10, LSUN, ImageNet), both in the cases when a perturbation layer is adopted only for Generators and when it is introduced to both Generator and Discriminator.