Unifying GANs and Score-Based Diffusion as Generative Particle Models

TL;DR

This paper presents a unified framework that connects particle-based diffusion models and GANs, showing that generator training can be viewed as a generalization of particle models, enabling new hybrid models.

Contribution

It introduces a novel unifying framework that treats generator training as a form of particle model, allowing integration of generators into diffusion models and GANs without generators.

Findings

Successfully integrated generator into score-based diffusion models.

Demonstrated GAN training without a generator.

Provided empirical evidence for the framework's viability.

Abstract

Particle-based deep generative models, such as gradient flows and score-based diffusion models, have recently gained traction thanks to their striking performance. Their principle of displacing particle distributions using differential equations is conventionally seen as opposed to the previously widespread generative adversarial networks (GANs), which involve training a pushforward generator network. In this paper we challenge this interpretation, and propose a novel framework that unifies particle and adversarial generative models by framing generator training as a generalization of particle models. This suggests that a generator is an optional addition to any such generative model. Consequently, integrating a generator into a score-based diffusion model and training a GAN without a generator naturally emerge from our framework. We empirically test the viability of these original models as proofs of concepts of potential applications of our framework.