b-GAN: Unified Framework of Generative Adversarial Networks

Masatosi Uehara, Issei Sato, Masahiro Suzuki, Kotaro Nakayama, Yutaka Matsuo

Abstract: Generative adversarial networks (GANs) are successful deep generative models. They are based on a two-player minimax game. However, the objective function derived in the original motivation is changed to obtain stronger gradients when learning the generator. We propose a novel algorithm that repeats density ratio estimation and f-divergence minimization. Our algorithm offers a new unified perspective toward understanding GANs and is able to make use of multiple viewpoints obtained from the density ratio estimation research, e.g. what divergence is stable and relative density ratio is useful.

TL;DR: New Unified Framework of Generative Adversarial Networks using Bregman divergence beyond f-GAN

Keywords: Deep learning, Unsupervised Learning

Conflicts: weblab.t.u-tokyo.ac.jp, k.u-tokyo.ac.jp, g.ecc.u-tokyo.ac.jp