Sequential Monte Carlo for Graphical Models

TL;DR

This paper introduces a novel framework employing sequential Monte Carlo algorithms for inference in probabilistic graphical models, enabling unbiased partition function estimation and high-dimensional sampling.

Contribution

It presents a new SMC-based approach for PGM inference, including unbiased partition function estimation and integration with particle MCMC for complex models.

Findings

Provides an unbiased estimate of the partition function.

Enables high-dimensional block sampling in PGMs.

Demonstrates effectiveness through theoretical analysis.

Abstract

We propose a new framework for how to use sequential Monte Carlo (SMC) algorithms for inference in probabilistic graphical models (PGM). Via a sequential decomposition of the PGM we find a sequence of auxiliary distributions defined on a monotonically increasing sequence of probability spaces. By targeting these auxiliary distributions using SMC we are able to approximate the full joint distribution defined by the PGM. One of the key merits of the SMC sampler is that it provides an unbiased estimate of the partition function of the model. We also show how it can be used within a particle Markov chain Monte Carlo framework in order to construct high-dimensional block-sampling algorithms for general PGMs.