Adversarial network training using higher-order moments in a modified Wasserstein distance

TL;DR

This paper introduces a generalized Wasserstein distance using higher-order moments for training GANs, leading to improved performance in generating complex data like antibody sequences.

Contribution

It proposes a novel higher-order Wasserstein metric for GAN training, enhancing data generation quality over traditional methods.

Findings

Higher-order Wasserstein improves mode coverage

GANs trained with this metric produce more realistic data

Enhanced performance demonstrated in antibody sequence generation

Abstract

Generative-adversarial networks (GANs) have been used to produce data closely resembling example data in a compressed, latent space that is close to sufficient for reconstruction in the original vector space. The Wasserstein metric has been used as an alternative to binary cross-entropy, producing more numerically stable GANs with greater mode covering behavior. Here, a generalization of the Wasserstein distance, using higher-order moments than the mean, is derived. Training a GAN with this higher-order Wasserstein metric is demonstrated to exhibit superior performance, even when adjusted for slightly higher computational cost. This is illustrated generating synthetic antibody sequences.