BEGIN:VCALENDAR VERSION:2.0 X-WR-CALNAME:EventsCalendar PRODID:-//hacksw/handcal//NONSGML v1.0//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York LAST-MODIFIED:20240422T053451Z TZURL:https://www.tzurl.org/zoneinfo-outlook/America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZNAME:EDT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZNAME:EST TZOFFSETFROM:-0400 TZOFFSETTO:-0500 DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT CATEGORIES:College of Arts and Sciences,College of Engineering,Lectures and Seminars DESCRIPTION:Title: Building GP Surrogate Model with High-Dimensional Input Abstract: Gaussian process (GP) regression is a popular surrogate modeling tool for computer simulations in engineering and scientific domains. Howe ver, it often struggles with high computational costs and low prediction a ccuracy when the simulation involves too many input variables. In this tal k, I will present two different approaches to build Gaussian process surro gate model for experiments with high dimensional input. I first introduce an optimal kernel learning approach to identify the active variables, ther eby overcoming GP model limitations and enhancing system understanding. Th is method approximates the original GP model's covariance function through a convex combination of kernel functions, each utilizing low-dimensional subsets of input variables. The second approach is Bayesian bridge GP regr ession approach, in which we impose shrinkage penalty on the linear regres sion coefficients of the mean and correlation coefficients in the covarian ce function. This is equivalent to using certain proper informative priors on these parameters under Bayesian framework. Using Spherical Hamiltonian Monte Carlo, we can directly sample from the constrained posterior distri bution without the restrictions on prior distribution as in Bayesian bridg e regression. Bio: Lulu Kang is a Professor in the Department of Mathemati cs and Statistics at the University of Massachusetts Amherst. She earned h er Ph.D. in Industrial Engineering and M.S. in Operations Research from th e Georgia Institute of Technology. Her research interests lie at the inter section of statistics and computational science, focusing on the design of experiments, uncertainty quantification, and statistical machine learning . Kang became an elected Fellow of the American Statistical Association in 2026. She has received the ASA's SPAIG Award and the SPES Outstanding Ser vice Award. She also serves as an Associate Editor for Technometrics and t he SIAM/ASA Journal on Uncertainty Quantification.\nEvent page: https://ww w.umassd.edu/events/cms/10-15-26-joint-dsccis-seminar-series-talk-by-dr-lu lu-kang-of-umass-amherst.php X-ALT-DESC;FMTTYPE=text/html:

Title: Building GP Surrogate Mo del with High-Dimensional Input

Abstract:

Gaussian process (GP) regression is a popular surrogate modeling tool for computer simulat ions in engineering and scientific domains. However\, it often struggles w ith high computational costs and low prediction accuracy when the simulati on involves too many input variables. In this talk\, I will present two di fferent approaches to build Gaussian process surrogate model for experimen ts with high dimensional input. I first introduce an optimal kernel learni ng approach to identify the active variables\, thereby overcoming GP model limitations and enhancing system understanding. This method approximates the original GP model's covariance function through a convex combination o f kernel functions\, each utilizing low-dimensional subsets of input varia bles. The second approach is Bayesian bridge GP regression approach\, in w hich we impose shrinkage penalty on the linear regression coefficients of the mean and correlation coefficients in the covariance function. This is equivalent to using certain proper informative priors on these parameters under Bayesian framework. Using Spherical Hamiltonian Monte Carlo\, we can directly sample from the constrained posterior distribution without the r estrictions on prior distribution as in Bayesian bridge regression.

\n< p>Bio:

Lulu Kang is a Professor in the Department of Mathematics a nd Statistics at the University of Massachusetts Amherst. She earned her P h.D. in Industrial Engineering and M.S. in Operations Research from the Ge orgia Institute of Technology. Her research interests lie at the intersect ion of statistics and computational science\, focusing on the design of ex periments\, uncertainty quantification\, and statistical machine learning. Kang became an elected Fellow of the American Statistical Association in 2026. She has received the ASA's SPAIG Award and the SPES Outstanding Serv ice Award. She also serves as an Associate Editor for Technometrics and th e SIAM/ASA Journal on Uncertainty Quantification.

Event page: https://www.umassd.edu/events/c ms/10-15-26-joint-dsccis-seminar-series-talk-by-dr-lulu-kang-of-umass-amhe rst.php

DTSTAMP:20260928T160931 DTSTART;TZID=America/New_York:20261015T133000 DTEND;TZID=America/New_York:20261015T150000 LOCATION:CCB242 SUMMARY;LANGUAGE=en-us:Joint DSC/CIS seminar series (Talk by Dr. Lulu Kang of UMass Amherst) UID:6184e047ca28fd0ef7d7fef3d7a8ab37@www.umassd.edu END:VEVENT END:VCALENDAR