Forward-KL Convergence of Time-Inhomogeneous Langevin Diffusions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Habring, Andreas, Zach, Martin
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909032490467328
author Habring, Andreas
Zach, Martin
author_facet Habring, Andreas
Zach, Martin
contents Many practical samplers rely on time-dependent drifts -- often induced by annealing or tempering schedules -- to improve exploration and stability. This motivates a unified non-asymptotic analysis of the corresponding Langevin diffusions and their discretizations. We provide a convergence analysis that includes non-asymptotic bounds for the continuous-time diffusion and its Euler--Maruyama discretization in the forward-Kullback--Leibler divergence under a single set of abstract conditions on the time-dependent drift. The results apply to many practically-relevant annealing schemes, including geometric tempering and annealed Langevin sampling. In addition, we provide numerical experiments comparing the annealing schemes covered by our theory in low- and high-dimensional settings.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22349
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Forward-KL Convergence of Time-Inhomogeneous Langevin Diffusions
Habring, Andreas
Zach, Martin
Numerical Analysis
Optimization and Control
65C40, 65C05, 68U10, 65C60
G.3; I.4.5
Many practical samplers rely on time-dependent drifts -- often induced by annealing or tempering schedules -- to improve exploration and stability. This motivates a unified non-asymptotic analysis of the corresponding Langevin diffusions and their discretizations. We provide a convergence analysis that includes non-asymptotic bounds for the continuous-time diffusion and its Euler--Maruyama discretization in the forward-Kullback--Leibler divergence under a single set of abstract conditions on the time-dependent drift. The results apply to many practically-relevant annealing schemes, including geometric tempering and annealed Langevin sampling. In addition, we provide numerical experiments comparing the annealing schemes covered by our theory in low- and high-dimensional settings.
title Forward-KL Convergence of Time-Inhomogeneous Langevin Diffusions
topic Numerical Analysis
Optimization and Control
65C40, 65C05, 68U10, 65C60
G.3; I.4.5
url https://arxiv.org/abs/2601.22349