Fit a GP emulator to a log-likelihood ensemble — gp_fit • rmorie

Wilkinson (2014) Sec. 2.1: quadratic mean, conjugate improper \(1/tau^2\) prior on \((beta, tau^2)\), plug-in MLE length-scale.

Usage

gp_fit(
  design,
  values,
  nugget = NULL,
  lengthscale = NULL,
  kernel = "sqexp",
  tau2 = NULL
)

Arguments

design: See Usage.
values: See Usage.
nugget: See Usage.
lengthscale: See Usage.
kernel: See Usage.
tau2: See Usage.

Examples

fit <- gp_fit(design = seq(0, 1, length.out = 6), values = cos(3 * seq(0, 1, length.out = 6)))
names(fit)
#>  [1] "design"       "values"       "beta"         "tau2"         "lengthscale" 
#>  [6] "kernel"       "nugget"       "chol"         "Ainv_r"       "Ainv_H"      
#> [11] "H"            "HtAinvH_chol" "n"            "q"            "dim"