Spectral Normalization for Generative Adversarial Networks

TL;DR

This paper introduces spectral normalization, a simple and efficient technique to stabilize GAN training by controlling the discriminator's spectral norm, leading to improved image quality across multiple datasets.

Contribution

The paper proposes spectral normalization as a novel weight normalization method for stabilizing GAN training, which is easy to implement and computationally efficient.

Findings

Spectral normalization improves GAN training stability.

SN-GANs generate higher quality images on benchmark datasets.

The method is easy to incorporate into existing GAN architectures.

Abstract

One of the challenges in the study of generative adversarial networks is the instability of its training. In this paper, we propose a novel weight normalization technique called spectral normalization to stabilize the training of the discriminator. Our new normalization technique is computationally light and easy to incorporate into existing implementations. We tested the efficacy of spectral normalization on CIFAR10, STL-10, and ILSVRC2012 dataset, and we experimentally confirmed that spectrally normalized GANs (SN-GANs) is capable of generating images of better or equal quality relative to the previous training stabilization techniques.