Saved in:
Bibliographic Details
Main Authors: Fu, Guoji, Suzuki, Taiji, Lee, Wee Sun, Nitanda, Atsushi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.15822
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911688258748416
author Fu, Guoji
Suzuki, Taiji
Lee, Wee Sun
Nitanda, Atsushi
author_facet Fu, Guoji
Suzuki, Taiji
Lee, Wee Sun
Nitanda, Atsushi
contents Score-based generative models are trained in high-dimensional ambient spaces, yet many data distributions are supported on low-dimensional nonlinear structures. We prove that, for compact $d$-dimensional smooth manifolds $\mathcal{M} \subset [0,1]^D$ with $d > 2$ and $β$-Hölder densities strictly positive on $\mathcal{M}$, a variance-preserving SGM estimator attains the intrinsic Wasserstein--1 sample exponent $\tilde{\mathcal{O}}(D^{\mathcal{O}_β(d)}n^{-(β+1)/(d+2β)})$, up to logarithmic factors and explicit geometry and density factors. The full nonasymptotic bound explicitly isolates the finite-order geometry envelope, Hölder radius, density lower bound, ambient dependence, and finite-order correction terms. The analysis separates score approximation into a large-noise tangent-cell regime and a small-noise projection-centered, de-Gaussianized Laplace regime. The key technical ingredient is a ReLU implementation of nearest-projection coordinates via finite intrinsic anchors and Gauss--Newton iterations, rather than approximating the manifold projection as a black-box high-dimensional smooth map. Consequently, for families with polynomially controlled geometry and density lower bounds, the constructed score-network parameters have polynomial ambient dependence.
format Preprint
id arxiv_https___arxiv_org_abs_2605_15822
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Intrinsic Wasserstein Rates for Score-Based Generative Models on Smooth Manifolds
Fu, Guoji
Suzuki, Taiji
Lee, Wee Sun
Nitanda, Atsushi
Machine Learning
Score-based generative models are trained in high-dimensional ambient spaces, yet many data distributions are supported on low-dimensional nonlinear structures. We prove that, for compact $d$-dimensional smooth manifolds $\mathcal{M} \subset [0,1]^D$ with $d > 2$ and $β$-Hölder densities strictly positive on $\mathcal{M}$, a variance-preserving SGM estimator attains the intrinsic Wasserstein--1 sample exponent $\tilde{\mathcal{O}}(D^{\mathcal{O}_β(d)}n^{-(β+1)/(d+2β)})$, up to logarithmic factors and explicit geometry and density factors. The full nonasymptotic bound explicitly isolates the finite-order geometry envelope, Hölder radius, density lower bound, ambient dependence, and finite-order correction terms. The analysis separates score approximation into a large-noise tangent-cell regime and a small-noise projection-centered, de-Gaussianized Laplace regime. The key technical ingredient is a ReLU implementation of nearest-projection coordinates via finite intrinsic anchors and Gauss--Newton iterations, rather than approximating the manifold projection as a black-box high-dimensional smooth map. Consequently, for families with polynomially controlled geometry and density lower bounds, the constructed score-network parameters have polynomial ambient dependence.
title Intrinsic Wasserstein Rates for Score-Based Generative Models on Smooth Manifolds
topic Machine Learning
url https://arxiv.org/abs/2605.15822