2013 Academic Year Seminars
Speaker(s): Dan Cornford
In this talk I will address the issue of how to incorporate prior knowledge in the form of complex computer simulators in Bayesian analysis. I will start by providing a rationale for a Bayesian approach to treating complex (natural) systems. I will introduce the concept of emulation, which is the approximation of computer simulators by statistical models. I will introduce the Managing Uncertainty in Complex Models project and highlight the role that Gaussian processes play in emulation. I will then discuss two particular aspects of emulation, namely the emulation of dynamic and stochastic simulators. I will show what progress has been made, and suggest the remaining open questions and speculate on possible solutions.