Compute GP gradient — gradientGP • bandle

Internal R function to pass R to C++, not for external use.

Function to perform Metropolis-Hastings for GP hyperparameters with different priors

gradientGP(Xk, tau, h, nk, D)

gradientGPmatern(Xk, tau, h, nk, D, materncov, nu)

posteriorgradientGPmatern(Xk, tau, h, nk, D, materncov, nu, hyppar)

gradientlogprior(h, hyppar)

likelihoodGP(Xk, tau, h, nk, D)

likelihoodGPmatern(Xk, tau, h, nk, D, materncov, nu)

posteriorGPmatern(Xk, tau, h, nk, D, materncov, nu, hyppar)

Gumbel(x, lambda, log = TRUE)

PCrhomvar(rho, a, lambda1, lambda2, log = TRUE)

metropolisGP(
  inith,
  X,
  tau,
  nk,
  D,
  niter,
  hyperMean = c(0, 0, 0),
  hyperSd = c(1, 1, 1)
)

metropolisGPmatern(
  inith,
  X,
  tau,
  nk,
  D,
  niter,
  nu = 2,
  hyppar = c(1, 1, 1),
  propSd = c(0.3, 0.1, 0.1)
)

Gumbel(x, lambda, log = TRUE)

PCrhomvar(rho, a, lambda1, lambda2, log = TRUE)

Arguments

Xk: The data
tau: The indexing parameters
h: GP hyperparameters
nk: Number of observations
D: number of samples
materncov: logical indicating whether matern covariance is used
nu: Smoothness of the matern covariance
hyppar: A vector indicating the penalised complexity prior hyperparameters. Default is c(1,1,1)
x: observation
lambda: scale parameter of the type-2 Gumbel distribution
log: logical indicating whether to return log. Default is TRUE
rho: length-scale parameter
a: amplitude
lambda1: first parameter of distribution
lambda2: second parameter of distribution
inith: initial hyperparamters
X: The data
niter: Number of MH iteractions
hyperMean: A vector indicating the log-normal means. Default is c(0,0,0).
hyperSd: A vector indicating the log-normal standard deviations. Default is c(1,1,1)
propSd: The proposal standard deviation. Default is c(0.3,0.1,0.1). Do not change unless you know what you are doing.

Value

Returns gp gradient

Returns the gradient of the posterior

return the gradient of the log prior, length-scale, aamplitude and noise

Returns gp negative log likelihood

Returns the negative log posterior of the GP

Returns the likelihood of the type-2 GUmbel distribution

Returns the likelihood of the bivariate penalised complexity prior

Returns new hyperparamters and the acceptance rate

Returns the likelihood of the type-2 GUmbel distribution

Returns the likelihood of the bivariate penalised complexity prior

Examples

Gumbel(3, lambda = 1)
#> [1] -2.918416