mgp {gptools2}

R Documentation

Multivariate Gaussian Process

Description

Multivariate Gaussian Process

Usage

mgp(X, Y, kernel = squared_exponential(), Sigma, ...)

Arguments

X	A numeric matrix; the data.
Y	A numeric matrix; the data.
kernel	A kernel object, which is a named list of a list of named parameters and an uninitialised kernel function.
Sigma	A numeric matrix; the covariance of the noise of the data. The dimension of 'Sigma' should be equal to the dimension of 'Y'.
...	Optional argument to pass to optim.

Examples

## Not run: 
X <- matrix(10 * runif(300), ncol = 3)
f <- function(X) {
  Y <- matrix(0, nrow = nrow(X), ncol = 2)
  Y[,1] <- sin(X[,1] + X[,2]) + 2 * X[,3]
  Y[,2] <- cos(X[,1] - X[,3]) + X[,2]
  Y
}
Y <- f(X)

model <- mgp(X, Y, Sigma = diag(2))
fitted_Y <- predict_mgp(model, X)$mean
head(cbind(Y, fitted_Y))
compare(as.numeric(Y), as.numeric(fitted_Y))

new_X <- matrix(10 * runif(90), ncol = 3)
new_Y <- f(new_X)
fitted_new_Y <- predict_mgp(model, new_X)$mean
head(cbind(new_Y, fitted_new_Y))
compare(as.numeric(new_Y), as.numeric(fitted_new_Y))

## End(Not run)

---