Saved in:
Bibliographic Details
Main Authors: Ilin, Vasily, Sushko, Peter, Hu, Jingwei
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.18130
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915563939299328
author Ilin, Vasily
Sushko, Peter
Hu, Jingwei
author_facet Ilin, Vasily
Sushko, Peter
Hu, Jingwei
contents We propose a deterministic sampling framework using Score-Based Transport Modeling for sampling an unnormalized target density $π$ given only its score $\nabla \log π$. Our method approximates the Wasserstein gradient flow on $\mathrm{KL}(f_t\|π)$ by learning the time-varying score $\nabla \log f_t$ on the fly using score matching. While having the same marginal distribution as Langevin dynamics, our method produces smooth deterministic trajectories, resulting in monotone noise-free convergence. We prove that our method dissipates relative entropy at the same rate as the exact gradient flow, provided sufficient training. Numerical experiments validate our theoretical findings: our method converges at the optimal rate, has smooth trajectories, and is often more sample efficient than its stochastic counterpart. Experiments on high-dimensional image data show that our method produces high-quality generations in as few as 15 steps and exhibits natural exploratory behavior. The memory and runtime scale linearly in the sample size.
format Preprint
id arxiv_https___arxiv_org_abs_2504_18130
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Score-based deterministic density sampling
Ilin, Vasily
Sushko, Peter
Hu, Jingwei
Machine Learning
Probability
Statistics Theory
Primary 65C05, 35Q84, 49Q22, Secondary 60H10, 68T07
We propose a deterministic sampling framework using Score-Based Transport Modeling for sampling an unnormalized target density $π$ given only its score $\nabla \log π$. Our method approximates the Wasserstein gradient flow on $\mathrm{KL}(f_t\|π)$ by learning the time-varying score $\nabla \log f_t$ on the fly using score matching. While having the same marginal distribution as Langevin dynamics, our method produces smooth deterministic trajectories, resulting in monotone noise-free convergence. We prove that our method dissipates relative entropy at the same rate as the exact gradient flow, provided sufficient training. Numerical experiments validate our theoretical findings: our method converges at the optimal rate, has smooth trajectories, and is often more sample efficient than its stochastic counterpart. Experiments on high-dimensional image data show that our method produces high-quality generations in as few as 15 steps and exhibits natural exploratory behavior. The memory and runtime scale linearly in the sample size.
title Score-based deterministic density sampling
topic Machine Learning
Probability
Statistics Theory
Primary 65C05, 35Q84, 49Q22, Secondary 60H10, 68T07
url https://arxiv.org/abs/2504.18130