kmMonotonic1D: Kriging model with monotonicity 1D constraint

In maatouk/constrKriging: Kriging with inequality constraints

Kriging model with monotonicity 1D constraint

Usage

kmMonotonic1D(design, response, basis.size = dim(design)[1] + 2 + 10,
  covtype = "matern5_2", basis.type = "C1", coef.cov = 0.5 * (max(design)
  - min(design)), coef.var = var(response), nugget = 1e-07 * sd(response))

Arguments

design	1-column matrix of the design of experiments
response	a vector containing the output values given by the real function at the design points
basis.size	a value represents the number of the basis functions (descritization of 1D input set)
covtype	an optimal character string specifying the covariance function to be used ("gauss" and "matern3_2" choice)
basis.type	an optimal character string specifying the regularization of the basis functions to be used (which implies the class of the simulation paths)
coef.cov	a value corresponding to the length theta hyper-parameters of covariance function
coef.var	a value specifying the variance parameter
nugget	an optimal value used as nugget effect to solve the numerical inverse matrix problem

Examples

kmMonotonic1D(design=c(0.1, 0.5, 0.9), response=c(1, 5, 9))
kmMonotonic1D(design=c(0.1, 0.5, 0.9), response=c(1, 5, 9), basis.type="C0")
kmMonotonic1D(design=sort(runif(5)), response=cumsum(runif(5)), nugget=1e-4, covtype="gauss")
f <- function(x){
log(20*x+1)
}
design <- c(0, 0.1, 0.2, 0.3, 0.4, 0.9, 1)
response <- f(design)
meany <- mean(response)
f <- function(x){
log(20*x+1)-meany
}
design <- c(0, 0.1, 0.2, 0.3, 0.4, 0.9, 1)
response <- f(design)
model = kmMonotonic1D(design, response, coef.var=335^2, coef.cov=4.7,basis.size=50)

maatouk/constrKriging documentation built on April 24, 2024, 7:13 p.m.