drawkpc: Draw kernel pca coefficients

In rdetools: Relevant Dimension Estimation (RDE) in Feature Spaces

Description

The function plots the absolute values of the kernel pca coefficients. The estimated relevant dimension and the estimated noise level (if available) are also drawn. Optionally, it puts a rescaled version of the loo-cv-error/negative-log-likelihood into the plot.

Usage

drawkpc(model,
        err = TRUE,
	pointcol = "blue",
	rdcol = "red",
	noisecol = "black",
	errcol = "brown",
	...)

Arguments

model	list of rde data returned by rde or selectmodel
err	leave this TRUE, if you want to have a rescaled version of the the loo-cv-error/negative-log-likelihood in the plot
pointcol	color of the kernel pca coefficients
rdcol	color of the relevant dimension line
noisecol	color of the noise level line
errcol	color of the the loo-cv-error/negative-log-likelihood
...	additional parameters to the plotting functions

Author(s)

Jan Saputra Mueller

References

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

See Also

rde, selectmodel, modelimage, distimage

Examples

## draw kernel pca coefficients after calling rde
d <- sincdata(100, 0.1) # generate sinc data
K <- rbfkernel(d$X)
r <- rde(K, d$y, est_noise = TRUE)
drawkpc(r)

## draw kernel pca coefficients after calling selectmodel
d <- sincdata(100, 0.1) # generate sinc data
m <- selectmodel(d$X, d$y, est_noise = TRUE, sigma = logspace(-3, 3, 100))
drawkpc(m)

Example output

rdetools documentation built on May 2, 2019, 7:02 a.m.