Hybrid Markov Chain Monte Carlo Using Gaussian Processes

Hybrid Markov chain Monte Carlo (MCMC) for sampling from multimodal target distributions when derivatives are unavailable. A Gaussian process approximation is used to emulate derivatives, enabling efficient exploration with parallel tempering. The method is described in Fielding, Nott and Liong (2011) . The research was carried out as part of the Singapore-Delft Water Alliance Multi-Objective Multi-Reservoir Management programme (R-264-001-272).