spectralGP (version 1.3.3)

xy2unit: Scales locations to the unit hypercube for use in spectral GP

Description

Scales locations to $(0,1)^d$ so that they can be related to the gridpoints in a spectral GP representation. The locations.scale argument allows one to scale the locations to a separate set of locations. E.g., if one wants to predict over a certain set of locations, but has a separate training set of locations that lie within the prediction set, one would use the prediction locations as the locations.scale argument.

Usage

xy2unit(locations, locations.scale = NULL)

Arguments

locations

A two-column matrix-like object (vector for one-dimensional data) of locations to be scaled.

locations.scale

A two-column matrix-like object (vector for one-dimensional data) of locations that provides the function with the min and max coordinates in each direction.

Value

A matrix (vector for one-dimensional data) of scaled locations lying in $(0,1)^d$.

Details

One may want to use both training and prediction locations as the locations.scale argument to ensure that all locations of interest will lie in $(0,1)^d$ and be able to be related to the gridpoints.

References

Type 'citation("spectralGP")' for references.

See Also

gp, new.mapping

Examples

library(spectralGP)
gp1=gp(c(128,128),matern.specdens,c(1,4))
n=100
locs=cbind(runif(n,0.2,1.2),runif(n,-0.2,1.4))
locs.predict=cbind(runif(n,-0.4,0.8),runif(n,-0.1,1.7))
scaled.locs=xy2unit(locs,rbind(locs,locs.predict))
scaled.locs.predict=xy2unit(locs.predict,rbind(locs,locs.predict))
train.map=new.mapping(gp1,scaled.locs)
predict.map=new.mapping(gp1,scaled.locs.predict)
plot(locs,xlim=c(min(locs[,1],locs.predict[,1]),max(locs[,1],
  locs.predict[,1])),ylim=c(min(locs[,2],locs.predict[,2]),
  max(locs[,2],locs.predict[,2])))
points(locs.predict,col=2)
plot(scaled.locs,xlim=c(0,1),ylim=c(0,1))
points(scaled.locs.predict,col=2)