Saved in:
Bibliographic Details
Main Authors: Li, Gen, Zhou, Yuchen, Wei, Yuting, Chen, Yuxin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.24042
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912535260692480
author Li, Gen
Zhou, Yuchen
Wei, Yuting
Chen, Yuxin
author_facet Li, Gen
Zhou, Yuchen
Wei, Yuting
Chen, Yuxin
contents In this paper, we explore provable acceleration of diffusion models without any additional retraining. Focusing on the task of approximating a target data distribution in $\mathbb{R}^d$ to within $\varepsilon$ total-variation distance, we propose a principled, training-free sampling algorithm that requires only the order of $$ d^{1+2/K} \varepsilon^{-1/K} $$ score function evaluations (up to log factor) in the presence of accurate scores, where $K>0$ is an arbitrary fixed integer. This result applies to a broad class of target data distributions, without the need for assumptions such as smoothness or log-concavity. Our theory is robust vis-a-vis inexact score estimation, degrading gracefully as the score estimation error increases -- without demanding higher-order smoothness on the score estimates as assumed in previous work. The proposed algorithm draws insight from high-order ODE solvers, leveraging high-order Lagrange interpolation and successive refinement to approximate the integral derived from the probability flow ODE. More broadly, our work develops a theoretical framework towards understanding the efficacy of high-order methods for accelerated sampling.
format Preprint
id arxiv_https___arxiv_org_abs_2506_24042
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Faster Diffusion Models via Higher-Order Approximation
Li, Gen
Zhou, Yuchen
Wei, Yuting
Chen, Yuxin
Machine Learning
Numerical Analysis
Statistics Theory
In this paper, we explore provable acceleration of diffusion models without any additional retraining. Focusing on the task of approximating a target data distribution in $\mathbb{R}^d$ to within $\varepsilon$ total-variation distance, we propose a principled, training-free sampling algorithm that requires only the order of $$ d^{1+2/K} \varepsilon^{-1/K} $$ score function evaluations (up to log factor) in the presence of accurate scores, where $K>0$ is an arbitrary fixed integer. This result applies to a broad class of target data distributions, without the need for assumptions such as smoothness or log-concavity. Our theory is robust vis-a-vis inexact score estimation, degrading gracefully as the score estimation error increases -- without demanding higher-order smoothness on the score estimates as assumed in previous work. The proposed algorithm draws insight from high-order ODE solvers, leveraging high-order Lagrange interpolation and successive refinement to approximate the integral derived from the probability flow ODE. More broadly, our work develops a theoretical framework towards understanding the efficacy of high-order methods for accelerated sampling.
title Faster Diffusion Models via Higher-Order Approximation
topic Machine Learning
Numerical Analysis
Statistics Theory
url https://arxiv.org/abs/2506.24042