A Unifying Generator Loss Function for Generative Adversarial Networks

Abstract

A unifying -parametrized generator loss function is introduced for a dual-objective generative adversarial network (GAN), which uses a canonical (or classical) discriminator loss function such as the one in the original GAN (VanillaGAN) system. The generator loss function is based on a symmetric class probability estimation type function, L_, and the resulting GAN system is termed L_-GAN. Under an optimal discriminator, it is shown that the generator's optimization problem consists of minimizing a Jensen-f_-divergence, a natural generalization of the Jensen-Shannon divergence, where f_ is a convex function expressed in terms of the loss function L_. It is also demonstrated that this L_-GAN problem recovers as special cases a number of GAN problems in the literature, including VanillaGAN, Least Squares GAN (LSGAN), Least kth order GAN (LkGAN) and the recently introduced (_D,_G)-GAN with _D=1. Finally, experimental results are conducted on three datasets, MNIST, CIFAR-10, and Stacked MNIST to illustrate the performance of various examples of the L_-GAN system.

Tasks

Reproductions