Software

bhetGP: Bayesian Heteroskedastic Gaussian Processes

This package fits a Bayesian Heteroskedastic Gaussian Process using Elliptical Slice Sampling (ESS) for inference of the latent noise process in conjunction with Metropolis-Hasting (MH) for kernel hyper-parameters. Additionally, the Woodbury-trick for efficient handling of replication is leveraged. For large data campaigns, Vecchia approximation is used. Methodology for this package is described in “Vecchia approximated Bayesian heteroskedastic Gaussian processes”