Minimax Optimality of Score-based Diffusion Models: Beyond the Density Lower Bound Assumptions

TL;DR

This paper establishes the asymptotic error bounds for score-based diffusion models in large-sample regimes, demonstrating near minimax optimality without requiring lower bound assumptions on the data distribution.

Contribution

It proves the optimality of kernel-based score estimators and diffusion models under minimal assumptions, extending previous results to broader nonparametric settings.

Findings

Kernel-based score estimator achieves optimal mean square error.

Diffusion model attains near minimax optimality under sub-Gaussian and Sobolev conditions.

Total variation error bound of the generated distribution is established.

Abstract

We study the asymptotic error of score-based diffusion model sampling in large-sample scenarios from a non-parametric statistics perspective. We show that a kernel-based score estimator achieves an optimal mean square error of $\widetilde{O}\left(n^{-1} t^{-\frac{d+2}{2}}(t^{\frac{d}{2}} \vee 1)\right)$ for the score function of $p_0*\mathcal{N}(0,t\boldsymbol{I}_d)$, where $n$ and $d$ represent the sample size and the dimension, $t$ is bounded above and below by polynomials of $n$, and $p_0$ is an arbitrary sub-Gaussian distribution. As a consequence, this yields an $\widetilde{O}\left(n^{-1/2} t^{-\frac{d}{4}}\right)$ upper bound for the total variation error of the distribution of the sample generated by the diffusion model under a mere sub-Gaussian assumption. If in addition, $p_0$ belongs to the nonparametric family of the $\beta$-Sobolev space with $\beta\le 2$, by adopting an early stopping strategy, we obtain that the diffusion model is nearly (up to log factors) minimax optimal. This removes the crucial lower bound assumption on $p_0$ in previous proofs of the minimax optimality of the diffusion model for nonparametric families.