Convergence Bounds for Sequential Monte Carlo on Multimodal Distributions using Soft Decomposition

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lee, Holden, Santana-Gijzen, Matheau
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914400445661184
author Lee, Holden
Santana-Gijzen, Matheau
author_facet Lee, Holden
Santana-Gijzen, Matheau
contents We prove bounds on the variance of a function $f$ under the empirical measure of the samples obtained by the Sequential Monte Carlo (SMC) algorithm, with time complexity depending on local rather than global Markov chain mixing dynamics. SMC is a Markov Chain Monte Carlo (MCMC) method, which starts by drawing $N$ particles from a known distribution, and then, through a sequence of distributions, re-weights and re-samples the particles, at each instance applying a Markov chain for smoothing. In principle, SMC tries to alleviate problems from multi-modality. However, most theoretical guarantees for SMC are obtained by assuming global mixing time bounds, which are only efficient in the uni-modal setting. We show that bounds can be obtained in the truly multi-modal setting, with mixing times that depend only on local MCMC dynamics.
format Preprint
id arxiv_https___arxiv_org_abs_2405_19553
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Convergence Bounds for Sequential Monte Carlo on Multimodal Distributions using Soft Decomposition
Lee, Holden
Santana-Gijzen, Matheau
Statistics Theory
Machine Learning
Probability
We prove bounds on the variance of a function $f$ under the empirical measure of the samples obtained by the Sequential Monte Carlo (SMC) algorithm, with time complexity depending on local rather than global Markov chain mixing dynamics. SMC is a Markov Chain Monte Carlo (MCMC) method, which starts by drawing $N$ particles from a known distribution, and then, through a sequence of distributions, re-weights and re-samples the particles, at each instance applying a Markov chain for smoothing. In principle, SMC tries to alleviate problems from multi-modality. However, most theoretical guarantees for SMC are obtained by assuming global mixing time bounds, which are only efficient in the uni-modal setting. We show that bounds can be obtained in the truly multi-modal setting, with mixing times that depend only on local MCMC dynamics.
title Convergence Bounds for Sequential Monte Carlo on Multimodal Distributions using Soft Decomposition
topic Statistics Theory
Machine Learning
Probability
url https://arxiv.org/abs/2405.19553