bigdatadist (version 1.1)

rkhs: RKHS Representation

Description

This function allows you to fit discrete functional data (fdframe) as functions in RKHS solving a regularization problem as stated in Munoz (2010).

Usage

rkhs(fdframe, gamma=1, kerfunc='rbf', 
            kerpar=list(sigma=1, bias=0, degree=2))

Arguments

fdframe

functional data fdframe object.

gamma

regularization parameter.

kerfunc

kernel function rbf or poly to be used.

kerpar

a list of kernel parameters where sigma is the scale with both kernels.

Value

fdframe

raw data in an fdframe object.

f

estimated functional data

alpha

coefficients for the linear combination.

lambda.star

reduced coefficients for the linear combination.

References

Munoz A. et al (2010). Representing functional data using support vector machines. Pattern recognition letters, 31(6).

Examples

# NOT RUN {
data(merval); t <- as.Date(merval[1:100,1])
t <- as.numeric(( t - min(t) ) / 154)
raw.data <-fdframe(t = t, Y = merval[1:100,2:5])
plot(raw.data, xlab='time', ylab='Merval raw data')

f.data <- rkhs(raw.data, gamma = 0.001, kerpar = list(sigma = 8))

print(f.data)

plot(f.data, xlab='time', ylab='Merval data')

# }