ABSTRACT
This chapter focuses on Sequential Monte Carlo (SMC) and some related Markov chain Monte Carlo methods. It deals with a particular application as its main motivating example: inference in the context of discrete state-space hidden Markov models (HMM). The chapter outlines SMC methods are applicable much more generally, for example to general state-space HMMs and to fairly arbitrary models. It describes particle approximations of these inference techniques and discusses the general principles behind the class of methods and presents a few convergence results. The chapter explores how the ideas have been applied to a much wider class of graphical models than HMMs. It provides a number of modifications of the SMC methodology described so far, most of which are used in practice to reduce the asymptotic variance associated with the SMC approximations, or to extend the scope of the algorithm to almost arbitrary settings.
