Advances in Sequential Monte Carlo Methods ​

Across various domains, it is necessary to estimate sequentially the latent variables in a state-space model. In situations of non-linearity and/or non-Gaussianity, this can be computationally difficult. We review the recent advances in the class of computational methods which address the problem, known as Sequential Monte Carlo. We discuss their use in an on-line as well as an off-line setting, in which latter case their advantages are compared to those of traditional MCMC methods. ​