Robert B. Gramacy Professor of Statistics

Book on Surrogate Modeling

Research

I am a computational statistician. I specialize in areas of real-data analysis in the physical, engineering and biological sciences where the ideal modeling apparatus is impractical, or where the current solutions are inefficient and thus skimp on fidelity. That often requires new models, new methods, and new algorithms. My goal is to be impactful in all three areas while remaining grounded in the needs of a motivating application. I aim to release general purpose software for consumption by the scientific community at large, not only other statisticians.

One example comes from my Ph.D. work, now nearly twenty years ago, where I helped NASA design a computer experiment for a re-usable rocket booster. The software developed for this project, tgp for R, has since found wide applicability in areas as diverse as insurance, economics, climate science, epidemiology, and finance. My dissertation won three awards including the Savage Award for best thesis in applied Bayesian methodology. In 2017 I received a Facebook faculty award for my research on large scale surrogate modeling and Bayesian optimization.

Fast-forward to the recent past: I am excited about large scale surrogate modeling. One example is work surrounding R package called laGP which is aimed at big data regression, and computer model emulation, by local approximate Gaussian processes. The code in the package, which facilitates massive parallelization, has been used to tackle a large-scale computer model calibration problem arising in a radiative shock hydrodynamics, and for blackbox constrained optimization in a benchmark groundwater remediation exercise. Recent extensions have been developed to predict the atmospheric drag of satellites in orbit. This is important for positioning and collision detection. The University of Chicago Research Computing Center (RCC), to whom I am grateful for valuable high-performance computing (HPC) resources, did a puff piece on this work-in-progress.

Our work on heteroskedastic Gaussian processes and sequential design (package hetGP) with postdoc Mickaël Binois has been finding lots of applications involving stochastic simulation of epidemics and inventory. Dave Higdon and I have been advancing surrogate modeling methods for cosmology project on galaxy formation. I've been collaborating with folks at NASA on a next generation spacesuit. Finally, I'm excited about recent work with Dave and Annie Sauer Boottht on deep Gaussian process surrogates.

Finally, I am excited to be part of a team of researchers at VT, and the University of Florida, who have begun work on a recently funded NSF project to forecast phytoplankton blooms. I'll be looking for good students to help with that. The enterprise will involve many of the topics outlined above: large scale surrogate modeling of heteroskedastic computer models, and calibration of those surrogates to real data collected at a local reservoir.

Funding

My research has been funded by the following awards.

  • NSF URoL:ASC 2318861: applying rules of life to forecast emergent behavior of phytoplankton and advance water quality, with Cayelan Carey and others; see a write-up here.
  • NSF CDS&E 2152679: local Gaussian process approaches for predicting jump behaviors with Chiwoo Park at Florida State; see a write-up here.
  • Small sub-awards from Argonne for neurmorphic computing, and from NASA for active learning for extreme events.
  • NSF CDS&E-MMS 1521702 and 1821258: for optimal stochastic control, heteroskedastic Gaussian process regression, and design of computer experiments with applications to epidemiology and UAV tracking.
  • NSF CDS&E-MMS 1621746: extending local approximate Gaussian processes (laGP) as motivated by large scale response surface modeling of satellite drag and solar irradiance.
  • SciDAC; DOE Office of Science ASCR and High Energy Physics: new methods and algorithms in the area of extreme-scale inference and machine learning motivated by surrogate modeling and inverse problem challenges in cosmology.

Gramacy Lab Members

Folks listed below are current members. Interested in joining? Here is a guide to expectations and pointers for success for potential and current lab members.

  • Parul Patil, VT PhD student, 2024-pres
  • Steven Barnett, VT PhD student, 2023-pres
  • Anna Flowers, VT PhD student, 2023-pres
  • Andrew Cooper, VT PhD student, 2022-pres

Recent Alumni

More Alumni

Professional Service

I am a member of ISBA, ASA, IMS, INFORMS, SIAM, currently serve as treasurer for the International Society for Bayesian Analysis (ISBA), program chair for the ASA section on Statistics in Defense and National Security (SDNS) , and program chair elect for the ASA Section on Bayesian Statistical Science (SBSS).

From Jan 2023 I am serving as the the Editor in Chief at Technometrics. If you're new to Tech, or to stats publishing generally, you might be interested in my tutorial on How to write a Technometrics paper (2022).

I recently stepped down as a member of the editorial board for Technometrics, Annals of Applied Statistics, Journal of Uncertainty Quantification, and Bayesian Analysis, and was as guest editor for a Statistica Sinica special issue on Uncertainty Quantification. I was one of two moderators for arXiv stat.ME (statistical methodology). I serve on the management committee of the new Data Science in Science journal.