SMC samplers for Bayesian Optimisation and Discovery of Additive Kernel Structure

This paper proposes the use of Sequential Monte Carlo samplers (SMCs) in Bayesian Optimisation and the discovery of an additive kernel structure. We present a comparison between Markov chain Monte Carlo (MCMC) and SMCs in both settings. SMCs are capable of finding the structure of an unknown function well. SMC samplers show comparable results to MCMC in this application and have the advantage of not always requiring a burn-in period.