Neural Residual Diffusion Models for Deep Scalable Vision Generation

TL;DR

This paper introduces Neural Residual Diffusion Models, a scalable framework that enhances deep vision generation by incorporating learnable residual parameters, leading to state-of-the-art results in image and video synthesis.

Contribution

It proposes a novel neural residual architecture with gated residuals that align with diffusion dynamics, enabling scalable training and improved generative quality.

Findings

Achieves state-of-the-art scores on image and video benchmarks.

Demonstrates improved fidelity and consistency in generated content.

Supports large-scale scalable training of deep generative models.

Abstract

The most advanced diffusion models have recently adopted increasingly deep stacked networks (e.g., U-Net or Transformer) to promote the generative emergence capabilities of vision generation models similar to large language models (LLMs). However, progressively deeper stacked networks will intuitively cause numerical propagation errors and reduce noisy prediction capabilities on generative data, which hinders massively deep scalable training of vision generation models. In this paper, we first uncover the nature that neural networks being able to effectively perform generative denoising lies in the fact that the intrinsic residual unit has consistent dynamic property with the input signal's reverse diffusion process, thus supporting excellent generative abilities. Afterwards, we stand on the shoulders of two common types of deep stacked networks to propose a unified and massively scalable Neural Residual Diffusion Models framework (Neural-RDM for short), which is a simple yet meaningful change to the common architecture of deep generative networks by introducing a series of learnable gated residual parameters that conform to the generative dynamics. Experimental results on various generative tasks show that the proposed neural residual models obtain state-of-the-art scores on image's and video's generative benchmarks. Rigorous theoretical proofs and extensive experiments also demonstrate the advantages of this simple gated residual mechanism consistent with dynamic modeling in improving the fidelity and consistency of generated content and supporting large-scale scalable training. Code is available at https://github.com/Anonymous/Neural-RDM.