SDNS Webinar Series: Dr. Annie Sauer-Booth Deep Gaussian Process Surrogates for Computer Experiments

Deep Gaussian Process Surrogates for Computer Experiments 19 March 2024 Abstract: This talk provides an overview of Bayesian deep Gaussian processes (DGPs) as surrogate models for computer experiments. Computer experiments are invaluable tools for replacing and/or supplementing direct experimentation, particularly in settings where physical experimentation is restricted by ethical, time, financial, or practicality constraints. Such simulations are necessarily complex and require statistical “surrogate” models, trained on a limited budget of simulator evaluations, which can provide predictions and uncertainty quantification at untried input configurations. Gaussian process (GP) surrogates are the canonical choice, but they are limited by stationarity constraints. DGPs upgrade ordinary GPs through functional composition, in which intermediate GP layers warp the original inputs, providing flexibility to model non-stationary dynamics. In large data settings, we integrate Vecchia approximation for faster computation. In small data settings, we utilize strategic active learning/sequential designs with a variety of objectives including variance reduction, Bayesian optimization, and reliability analysis. We showcase implementation in the “deepgp” package for R on CRAN. About the Speaker: Dr. Annie Sauer-Booth is an Assistant Professor in the Department of Statistics at NC State University. Her research focuses on surrogate modeling of computer experiments including uncertainty quantification, active learning, Bayesian optimization, and reliability analysis. She completed her Ph.D. in statistics at Virginia Tech last year, where she worked with advisors Bobby Gramacy and Dave Higdon on developing deep Gaussian processes as surrogate models. 📖 TABLE OF CONTENTS 0:00:00 Welcome to the SDNS Webinar! 0:01:39 Beginning of Presentation 0:47:56 Questions and Closing Items American Statistical Association Section on Statistics in Defense and National Security https://community.amstat.org/sdns/home