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

Oliver Serang

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 i…