Title: Emulation, Adaptive Design, and Optimization Using Treed Gaussian Processes

Speaker: Herbert Lee, Univ. of California, Santa Cruz

Date/Time: Wednesday, January 26, 2011, 10:00 am (CA), 11:00 am (NM)       

Location: CSRI Building/Room 90 (Sandia NM)

Brief Abstract: Treed Gaussian process models are a computationally efficient way of capturing nonstationarity in a spatial process, such as the output of a computer model.  R code is available from CRAN in the library tgp. The Bayesian perspective allows full accounting for uncertainty, which is useful for designing sequential computer experiments, and also for optimization.  This talk will give a brief overview of the approach and provide examples of its use for modeling, adaptive design, optimization, and constrained optimization.

CSRI POC: Genetha Gray, 925-294-4957