Sequential Monte Carlo Methods - an introduction

Fanny Empacher (University of St Andrews, UK)

Sequential Monte Carlo (SMC) methods are a helpful tool for inference in State Space Models, in particular in cases where standard methods such as MCMC or the Kalman filter fail. SMC methods can handle non-linear and non-Gaussian models, and can be extended to suit many requirements. This talk and its continuation with Part II on February 26 will introduce these methods, discuss some theoretical properties and give example applications. I will also discuss typical problems and possible solutions, including some recent developments. Part I will focus on state inference methods which form the basis for parameter inference methods that will be discussed in Part II.