On the Effects of Batch and Weight Normalization in Generative Adversarial Networks

TL;DR

This paper introduces weight normalization (WN) for GAN training, demonstrating it improves stability, efficiency, and sample quality over batch normalization (BN), supported by a new evaluation metric and extensive experiments.

Contribution

The paper proposes a novel weight normalization method for GANs, providing a systematic evaluation framework and showing superior performance and stability compared to batch normalization.

Findings

WN reduces mean squared reconstruction error by 10%

WN achieves more stable training on deep ResNet architectures

WGAN produces higher quality samples than BN-based GANs

Abstract

Generative adversarial networks (GANs) are highly effective unsupervised learning frameworks that can generate very sharp data, even for data such as images with complex, highly multimodal distributions. However GANs are known to be very hard to train, suffering from problems such as mode collapse and disturbing visual artifacts. Batch normalization (BN) techniques have been introduced to address the training. Though BN accelerates the training in the beginning, our experiments show that the use of BN can be unstable and negatively impact the quality of the trained model. The evaluation of BN and numerous other recent schemes for improving GAN training is hindered by the lack of an effective objective quality measure for GAN models. To address these issues, we first introduce a weight normalization (WN) approach for GAN training that significantly improves the stability, efficiency and the quality of the generated samples. To allow a methodical evaluation, we introduce squared Euclidean reconstruction error on a test set as a new objective measure, to assess training performance in terms of speed, stability, and quality of generated samples. Our experiments with a standard DCGAN architecture on commonly used datasets (CelebA, LSUN bedroom, and CIFAR-10) indicate that training using WN is generally superior to BN for GANs, achieving 10% lower mean squared loss for reconstruction and significantly better qualitative results than BN. We further demonstrate the stability of WN on a 21-layer ResNet trained with the CelebA data set. The code for this paper is available at https://github.com/stormraiser/gan-weightnorm-resnet