Generative Adversarial Nets from a Density Ratio Estimation Perspective

TL;DR

This paper presents a new perspective on GANs by framing them as a process involving density ratio estimation and f-divergence minimization, offering insights into their stability and effectiveness.

Contribution

It introduces a novel algorithm that combines density ratio estimation with divergence minimization, providing a fresh understanding of GAN training dynamics.

Findings

The proposed method offers a new perspective on GANs.

It leverages multiple viewpoints from density ratio estimation research.

The approach enhances understanding of divergence stability in GANs.

Abstract

Generative adversarial networks (GANs) are successful deep generative models. GANs 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 the density ratio estimation and f-divergence minimization. Our algorithm offers a new perspective toward the understanding of GANs and is able to make use of multiple viewpoints obtained in the research of density ratio estimation, e.g. what divergence is stable and relative density ratio is useful.