Emulation of random output (stochastic) models: theory and application

Add to your list(s) Download to your calendar using vCal

• Dan Cornford, Non-linearity and Complexity Research Group, Aston University
• Tuesday 01 December 2009, 14:30-15:30
• Large Seminar Room, 1st Floor, Institute of Public Health, University Forvie Site, Robinson Way, Cambridge.

If you have a question about this talk, please contact Michael Sweeting.

In this talk we will briefly introduce the concept of emulation – the construction of a surrogate model – which is typically used to speed up (Bayesian) analysis using deterministic computer models, as presented in the MUCM project. Most existing approaches to emulation assume that the computer simulator is deterministic, however increasingly computer simulators are becoming stochastic. This typically arises from some unknown or random interactions within the simulator which induce random outputs. We will discuss how to build emulators of stochastic simulators, focusing on heteroscedastic modelling using Gaussian processes. In particular we will address the experimental design issue when developing stochastic emulators. We will present results from our new Fisher information based design approach, and show the application of this to a relatively simple two vector rabies model. We will conclude with a discussion of interesting future directions for building emulators for stochastic simulators.

MUCM : http://mucm.group.shef.ac.uk/index.html

This talk is part of the MRC Biostatistics Unit Seminars series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• Cambridge Immunology
• Large Seminar Room, 1st Floor, Institute of Public Health, University Forvie Site, Robinson Way, Cambridge
• Life Sciences
• Life Sciences
• MRC Biostatistics Unit Seminars

Note that ex-directory lists are not shown.