Dimension-Free Convergence of Diffusion Models for Approximate Gaussian Mixtures

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Gen, Cai, Changxiao, Wei, Yuting
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908305528455168
author Li, Gen
Cai, Changxiao
Wei, Yuting
author_facet Li, Gen
Cai, Changxiao
Wei, Yuting
contents Diffusion models are distinguished by their exceptional generative performance, particularly in producing high-quality samples through iterative denoising. While current theory suggests that the number of denoising steps required for accurate sample generation should scale linearly with data dimension, this does not reflect the practical efficiency of widely used algorithms like Denoising Diffusion Probabilistic Models (DDPMs). This paper investigates the effectiveness of diffusion models in sampling from complex high-dimensional distributions that can be well-approximated by Gaussian Mixture Models (GMMs). For these distributions, our main result shows that DDPM takes at most $\widetilde{O}(1/\varepsilon)$ iterations to attain an $\varepsilon$-accurate distribution in total variation (TV) distance, independent of both the ambient dimension $d$ and the number of components $K$, up to logarithmic factors. Furthermore, this result remains robust to score estimation errors. These findings highlight the remarkable effectiveness of diffusion models in high-dimensional settings given the universal approximation capability of GMMs, and provide theoretical insights into their practical success.
format Preprint
id arxiv_https___arxiv_org_abs_2504_05300
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Dimension-Free Convergence of Diffusion Models for Approximate Gaussian Mixtures
Li, Gen
Cai, Changxiao
Wei, Yuting
Machine Learning
Numerical Analysis
Statistics Theory
Diffusion models are distinguished by their exceptional generative performance, particularly in producing high-quality samples through iterative denoising. While current theory suggests that the number of denoising steps required for accurate sample generation should scale linearly with data dimension, this does not reflect the practical efficiency of widely used algorithms like Denoising Diffusion Probabilistic Models (DDPMs). This paper investigates the effectiveness of diffusion models in sampling from complex high-dimensional distributions that can be well-approximated by Gaussian Mixture Models (GMMs). For these distributions, our main result shows that DDPM takes at most $\widetilde{O}(1/\varepsilon)$ iterations to attain an $\varepsilon$-accurate distribution in total variation (TV) distance, independent of both the ambient dimension $d$ and the number of components $K$, up to logarithmic factors. Furthermore, this result remains robust to score estimation errors. These findings highlight the remarkable effectiveness of diffusion models in high-dimensional settings given the universal approximation capability of GMMs, and provide theoretical insights into their practical success.
title Dimension-Free Convergence of Diffusion Models for Approximate Gaussian Mixtures
topic Machine Learning
Numerical Analysis
Statistics Theory
url https://arxiv.org/abs/2504.05300