Emulating Stochastic Computer Models (and using Deterministic Models to do so).
Evan Baker (University of Exeter)
Friday 1st October 15:00-16:00 https://www.smartsurvey.co.uk/s/MNW32H/
Using fast statistical models (emulators) to predict the output of slower numerical models (simulators) is, at this point, a fairly well researched idea. It is usually assumed that these slow simulators are deterministic, with the simulation output always being the same if the same input settings are used. However, many simulators are in fact stochastic, with a random internal component leading to noisy outputs.
In this talk we will review some general strategies for dealing with these stochastic simulators, and some challenges these models raise. We will also investigate how deterministic simulators can be used to improve stochastic emulators, using a case study example involving the engineering design of buildings.