Join us for an afternoon with Serge Guillas.
Joyce Cummings Center, Room 280
April 5th 2024. 12-1pm
Title: Deep Gaussian Process emulation of simulators with multi-physics components and sharp transitions, with application to convection and climate simulations
Abstract: We first introduce Gaussian Process (GP) emulation of computer models. These are surrogates of simulators that efficiently mimic the input-output relationship of such complex numerical models, only sampling a small set of experiments. We illustrate the application of combining hybrid emulation of convection within a climate numerical model. We then present a new type of emulator of any feed forward multi-physics system, by linking GP emulators of individual simulators. The Deep Gaussian Process (DGP) is then presented as a surrogate that shares the structure of the linked emulator but enables the emulation of highly non-linear simulators without the knowledge of individual sub-processes. We then examine sharp changes in the Quantity of Interest (QoI) within computer simulators. These often indicate bifurcations or critical transitions within the investigated system, e.g. laminar v. turbulent behaviour in fluid dynamics. An efficient approach localises these changes with minimal number of evaluations is introduced. We demonstrate the efficacy of the proposed framework on the Rayleigh–Bénard convection.