Convergence of Langevin AIS for multimodal distributions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Agarwal, Akshat, Iyer, Gautam, Jameson, Aidan, Son, Seungjae, Wimmer, Wyatt
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914488863686656
author Agarwal, Akshat
Iyer, Gautam
Jameson, Aidan
Son, Seungjae
Wimmer, Wyatt
author_facet Agarwal, Akshat
Iyer, Gautam
Jameson, Aidan
Son, Seungjae
Wimmer, Wyatt
contents We study convergence rates of the annealed importance sampling algorithm (Neal '01) combined with Langevin Monte Carlo when the target is a multimodal Gibbs measure. The main result shows that for a fixed error threshold, the time complexity is quadratic in the inverse temperature. We identify a simple and useful quantity that controls the sampling error for AIS in a general setting, and then bound this quantity in our setting using spectral estimates. We also study an autonormalized version and obtain bounds for the time complexity in terms of the inverse temperature.
format Preprint
id arxiv_https___arxiv_org_abs_2604_17526
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Convergence of Langevin AIS for multimodal distributions
Agarwal, Akshat
Iyer, Gautam
Jameson, Aidan
Son, Seungjae
Wimmer, Wyatt
Probability
Statistics Theory
60J22, 65C05, 65C40
We study convergence rates of the annealed importance sampling algorithm (Neal '01) combined with Langevin Monte Carlo when the target is a multimodal Gibbs measure. The main result shows that for a fixed error threshold, the time complexity is quadratic in the inverse temperature. We identify a simple and useful quantity that controls the sampling error for AIS in a general setting, and then bound this quantity in our setting using spectral estimates. We also study an autonormalized version and obtain bounds for the time complexity in terms of the inverse temperature.
title Convergence of Langevin AIS for multimodal distributions
topic Probability
Statistics Theory
60J22, 65C05, 65C40
url https://arxiv.org/abs/2604.17526