splitpar: Split a parameter value within an interval

In mlesnoff/rnirs: Dimension reduction, Regression and Discrimination for Chemometrics

Split a parameter value within an interval

Description

Function splitpar split the value of a model parameter within a given range (interval). The function is a help for model tuning (e.g. RR, KRR, SVM etc.).

Usage

splitpar(x, a1 = .1, a2 = 10, n = 20)

Arguments

x	A scalar.
a1	The lower limit of the range interval is a1 * x.
a2	The upper limit of the range interval is a2 * x.
n	Approximate number of output values within the range.

Value

A vector (grid of parameter values).

Examples

lambda <- 1e-2
z <- splitpar(lambda, .1, 10, n = 20)
z
plot(z)

################## EXAMPLE OF TUNING

data(datcass)

Xr <- datcass$Xr
yr <- datcass$yr

Xu <- datcass$Xu
yu <- datcass$yu

Xr <- detrend(Xr)
Xu <- detrend(Xu)
dim(Xr)
dim(Xu)

### First round

lambda <- 10^(-15:5) ; unit <- 1
sigma <- 10^(-6:3)
fm <- krr(Xr, yr, Xu, yu, lambda = lambda, unit = unit, kern = krbf, sigma = sigma)
z <- mse(fm, ~ lambda + unit + sigma)
u <- z[z$rmsep == min(z$rmsep), ][1, ]
u
z$log.lambda <- log(z$lambda, base = 10)
group <- paste("sigma=", z$sigma, sep = "")
plotmse(z, namx = "log.lambda", group = group)

### Second round

lambda <- splitpar(u$lambda)
sigma <- splitpar(u$sigma)
fm <- krr(Xr, yr, Xu, yu, lambda = lambda, unit = unit, kern = krbf, sigma = sigma)
z <- mse(fm, ~ lambda + unit + sigma)
u <- z[z$rmsep == min(z$rmsep), ][1, ]
u

### Final estimate

fm <- krr(Xr, yr, Xu, yu, lambda = u$lambda, unit = u$unit, kern = krbf, sigma = u$sigma)
fit <- fm$fit$y1
r <- yu - fit
sqrt(mean(r^2))
plot(yu, fit)
abline(0, 1, col = "grey")

mlesnoff/rnirs documentation built on April 24, 2023, 4:17 a.m.