Statistics Seminar: Anthony Overstall
Anthony Overstall, St. Andrews University
Friday 28 March 2014, 1400-1500
A54, Postgraduate Statistics Centre Lecture Theatre
The approximate coordinate exchange algorithm for Bayesian optimal experimental design
Optimal Bayesian experimental design typically involves maximising an objective function which is the expectation of some appropriately chosen utility function with respect to the joint distribution of parameters and responses. This objective function is usually not available in closed form and the design space can be of high dimensionality. The approximate coordinate exchange algorithm is proposed to approach this maximisation problem where a Gaussian process emulator of Monte Carlo integration is used to approximate the objective function. The algorithm can be used for arbitrary utility functions including those that result in pseudo-Bayesian designs such as Bayesian D-optimal designs. If time allows an extension to the algorithm will be discussed that allows optimal Bayesian experimental design for calibration of computer models and, in particular models, derived from systems of non-linear differential equations.