GAN-QP: A Novel GAN Framework without Gradient Vanishing and Lipschitz Constraint

TL;DR

This paper introduces GAN-QP, a new GAN framework that avoids gradient vanishing and Lipschitz constraints, simplifying the training process and improving performance over WGAN.

Contribution

It proposes a novel GAN framework that bypasses the need for Lipschitz constraints by directly analyzing divergence properties in dual space.

Findings

GAN-QP outperforms WGAN in theory

GAN-QP demonstrates better practical performance

The divergence analysis simplifies GAN construction

Abstract

We know SGAN may have a risk of gradient vanishing. A significant improvement is WGAN, with the help of 1-Lipschitz constraint on discriminator to prevent from gradient vanishing. Is there any GAN having no gradient vanishing and no 1-Lipschitz constraint on discriminator? We do find one, called GAN-QP. To construct a new framework of Generative Adversarial Network (GAN) usually includes three steps: 1. choose a probability divergence; 2. convert it into a dual form; 3. play a min-max game. In this articles, we demonstrate that the first step is not necessary. We can analyse the property of divergence and even construct new divergence in dual space directly. As a reward, we obtain a simpler alternative of WGAN: GAN-QP. We demonstrate that GAN-QP have a better performance than WGAN in theory and practice.