kpr: Kernel projection regression

Description

The function does a kernel projection regression. It returns a function which predicts labels for new data points.

Usage

kpr(model, X = NULL, Xname = "X", Yname = "Y", kernel = NULL, regression = TRUE, ...)

Arguments

model

list of rde data returned by rde or selectmodel

matrix containing the data points, only needed if rde was used

Xname

the name of the parameter of the kernel function which should contain the data points, only needed if rde was used

Yname

the name of the parameter of the kernel function which should contain the 2nd data matrix

kernel

kernel function to use, only needed if rde was used

regression

set this to TRUE in case of a regression problem and to FALSE in case of a classification problem; only needed if rde was used

...

parameters for the kernel function, only needed if rde was used

Value

function which predicts labels for new input data (gets a matrix with one data point per line)

References

M. L. Braun, J. M. Buhmann, K. R. Mueller (2008) \_On Relevant Dimensions in Kernel Feature Spaces\_

Examples

Run this code

## kernel projection regression after
## calling selectmodel (recommended)
d <- sincdata(100, 0.1) # generate sinc data
# do model selection
m <- selectmodel(d$X, d$y, sigma = logspace(-3, 3, 100))
f <- kpr(m)
plot(f, -4, 4)