unscalingFun1d: Unscaling in one dimension

In IRSN/RobustInv: "Robust inversion of expensive black-box functions"

Description

Unscaling in one dimension using the knots of a km object

Usage

unscalingFun1d(y, knots, eta, standardize = FALSE, lower = NULL,
  upper = NULL)

Arguments

y	Array with values to be unscaled
knots	Array obtained from the field model@covariance@knots of a km object
eta	Array obtained from the field model@covariance@eta of a km object
standardize	If the initial values y are not in [0,S], with S the integral of the piecewise linear function equal to eta at points knots, then there is the possibility to rescale the values y by indicating in which interval they are.
lower	If standardize=TRUE, this is the lower bound of the interval where the yi's are.
upper	If standardize=TRUE, this is the upper bound of the interval where the yi's are.

Value

Array with size length(y) containing the scaled uncoordinates.

Author(s)

Clement Chevalier [e-mail]

Examples

library(DiceKriging)
knots <- c(1,2,3)
eta <- c(2,1,4)
t <- seq(from = 1, to = 3, length = 101)
scaled_t <- scalingFun1d(x = t,knots = knots,eta = eta)

result <- unscalingFun1d(scaled_t,knots=knots,eta=eta)
# now result is equal to t !

# an example of unscaling of uniformly distributed points to have more points in regions where eta is large.
myrands <- matrix( runif(2000),ncol=2  )

knots <- c(0,0.5,1)
eta <- c(5,1,5) # large on the bounds, low in the middle
res1 <- unscalingFun1d(y = myrands[,1] , knots=knots , eta = eta , standardize = TRUE, lower=0, upper = 1)
res2 <- unscalingFun1d(y = myrands[,2] , knots=knots , eta = eta , standardize = TRUE, lower=0, upper = 1)

plot(x=res1 , y=res2, type="p") # more points in the corners

IRSN/RobustInv documentation built on Nov. 20, 2019, 10:46 p.m.