An Example Of Sequential Monte Carlo Block Sampling Method

Sequential Monte Carlo methods are powerful algorithms to sample from sequences of complex probability distributions. They are mainly based on a combination of importance sampling and resampling techniques. The efficiency of these methods depends crucially on the sampling strategies adopted. In this paper, we present an extended importance sampling framework which allows more freedom than standard techniques to impute random samples. This makes it possible to develop efficient and original sampling strategies. Applications to optimal filtering problems illustrate this approach