deepgp Package

Maintainer: Annie Sauer anniees@vt.edu

Performs model fitting and sequential design for deep Gaussian processes following Sauer, Gramacy, and Higdon (2020). Models extend up to three layers deep; a one layer model is equivalent to typical Gaussian process regression. Sequential design criteria include integrated mean-squared error (IMSE), active learning Cohn (ALC), and expected improvement (EI). Covariance structure is based on inverse exponentiated squared euclidean distance. Applicable to noisy and deterministic functions. Incorporates SNOW parallelization and utilizes C under the hood.

View deepgp-package help file for more information.

Reference

Sauer, A, RB Gramacy, and D Higdon. 2020. “Active Learning for Deep Gaussian Process Surrogates.” arXiv:2012.08015.