Title: Bayesian History Matching of Complex Models
Abstract: In many scientific disciplines complex computer simulators are employed to help understand corresponding real-world physical processes. These models increasingly involve large numbers of unknown parameters, have a complex structure which can result in substantial evaluation time relative to the needs of the analysis, and need to be compared to observed data of various forms to generate reliable models to support decision-making. The correct analysis of such models usually demands a comprehensive exploration of the uncertain high dimensional input parameter space, that incorporates and respects the most important sources of uncertainty. This can be a difficult task, but it is essential for any meaningful inference or prediction to be made and hence represents a fundamental challenge. The challenge needs to be addressed considering the possible drawback of identifying only a partial subset of the input parameters consistent with observed data. Critically, these subsets will result in unjustifiable forecasts that are both biased and overconfident, which directly lead to sub-optimal decision making. A Bayesian statistical methodology is presented in order to identify the subset of the input space that could give rise to acceptable matches between model output and measured data. The approach involves the use of an iterative succession of emulators (stochastic belief specifications detailing beliefs about the complex model) embedded within a method to search high dimensional spaces while incorporating major sources of uncertainty.
Download the flyer here.