Convergence Analysis of Probability Flow ODE for Score-based Generative Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Huang, Daniel Zhengyu, Huang, Jiaoyang, Lin, Zhengjiang
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910914575335424
author Huang, Daniel Zhengyu
Huang, Jiaoyang
Lin, Zhengjiang
author_facet Huang, Daniel Zhengyu
Huang, Jiaoyang
Lin, Zhengjiang
contents Score-based generative models have emerged as a powerful approach for sampling high-dimensional probability distributions. Despite their effectiveness, their theoretical underpinnings remain relatively underdeveloped. In this work, we study the convergence properties of deterministic samplers based on probability flow ODEs from both theoretical and numerical perspectives. Assuming access to $L^2$-accurate estimates of the score function, we prove the total variation between the target and the generated data distributions can be bounded above by $\mathcal{O}(d^{3/4}δ^{1/2})$ in the continuous time level, where $d$ denotes the data dimension and $δ$ represents the $L^2$-score matching error. For practical implementations using a $p$-th order Runge-Kutta integrator with step size $h$, we establish error bounds of $\mathcal{O}(d^{3/4}δ^{1/2} + d\cdot(dh)^p)$ at the discrete level. Finally, we present numerical studies on problems up to 128 dimensions to verify our theory.
format Preprint
id arxiv_https___arxiv_org_abs_2404_09730
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Convergence Analysis of Probability Flow ODE for Score-based Generative Models
Huang, Daniel Zhengyu
Huang, Jiaoyang
Lin, Zhengjiang
Machine Learning
Numerical Analysis
Classical Analysis and ODEs
Score-based generative models have emerged as a powerful approach for sampling high-dimensional probability distributions. Despite their effectiveness, their theoretical underpinnings remain relatively underdeveloped. In this work, we study the convergence properties of deterministic samplers based on probability flow ODEs from both theoretical and numerical perspectives. Assuming access to $L^2$-accurate estimates of the score function, we prove the total variation between the target and the generated data distributions can be bounded above by $\mathcal{O}(d^{3/4}δ^{1/2})$ in the continuous time level, where $d$ denotes the data dimension and $δ$ represents the $L^2$-score matching error. For practical implementations using a $p$-th order Runge-Kutta integrator with step size $h$, we establish error bounds of $\mathcal{O}(d^{3/4}δ^{1/2} + d\cdot(dh)^p)$ at the discrete level. Finally, we present numerical studies on problems up to 128 dimensions to verify our theory.
title Convergence Analysis of Probability Flow ODE for Score-based Generative Models
topic Machine Learning
Numerical Analysis
Classical Analysis and ODEs
url https://arxiv.org/abs/2404.09730