Mar 9, 2017 - Seminar

Younus Al-taweel (Sheffield)
Diagnostics for Gaussian processes emulators
Hicks LT6, 2-3pm

Emulators are used to obtain approximations for computationally expensive computer models. Emulators can be constructed under a Bayesian framework using any joint probability distribution. The Gaussian distribution, however, is the simplest one that is mathematically tractable and leads to simple forms of the posterior mean and the posterior variance.

Emulators need to be subjected to a validation process since various assumptions have to be made. Diagnostic methods can be used for checking and validating Gaussian process emulators. However, it is not always possible to derive the required reference distribution for some diagnostics analytically. We develop a simulation-based method based on simulating samples from the posterior distribution of the output function. This simulation-based method can be used to obtain the reference distribution of diagnostics that cannot be obtained analytically. Moreover, we develop a coverage interval diagnostic that can be used to investigate whether the Gaussian process assumption is suitable or not.