The approximate coordinate exchange algorithm for Bayesian optimal design of experiments

Overstall, A. and Woods, D. (2015) The approximate coordinate exchange algorithm for Bayesian optimal design of experiments. Working Paper. University of Glasgow, Glasgow, UK. (Unpublished)




Optimal Bayesian experimental design typically involves maximising the expectation, with respect to the joint distribution of parameters and responses, of some appropriately chosen utility function. 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 for this maximisation problem where a Gaussian process emulator is used to approximate the objective function. The algorithm can be used for arbitrary utility functions meaning we can consider fully Bayesian optimal design. It can also be used for those utility functions that result in pseudo-Bayesian designs such as the popular Bayesian D-optimality. The algorithm is demonstrated on a range of examples.

Item Type:Research Reports or Papers (Working Paper)
Keywords:Bayesian coordinate exchange Gaussian process emulator optimal experimental design
Glasgow Author(s) Enlighten ID:Overstall, Dr Antony
Authors: Overstall, A., and Woods, D.
College/School:College of Science and Engineering > School of Mathematics and Statistics > Statistics
Publisher:University of Glasgow
Copyright Holders:Copyright © 2015 The Authors
Publisher Policy:Reproduced with the permission of the authors
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record