unscalingFun1d: Unscaling in one dimension
In IRSN/RobustInv: "Robust inversion of expensive black-box functions"

Description Usage Arguments Value Author(s) Examples

Unscaling in one dimension using the knots of a km object

1 2	unscalingFun1d(y, knots, eta, standardize = FALSE, lower = NULL, upper = NULL)

`y`	Array with values to be unscaled
`knots`	Array obtained from the field `[email protected]@knots` of a `km` object
`eta`	Array obtained from the field `[email protected]@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.

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

Clement Chevalier [email protected]

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