sampleClassif

Description

performs task analogous to mixKnn (i.e. leave-one-out classification), but uses synthetic representatives to infer labels, instead of k-NN. Each representative is obtained by concatenating all GMM (i.e. elements) of a specific label value, resampling from this redundant mixture, and applying varbayes on this sample.

Usage

1
sampleClassif(data, labels, KLparam = 500, rho = new.env())

Arguments

data

list of GMM.

labels

vector of numeric labels associated to data.

KLparam

number of samples for jsmc.

rho

R environment object. Used to issue R commands within the C routine.

Value

classification error ratio in [0,1].

Author(s)

Pierrick Bruneau

See Also

mixKnn

Examples

1
2
3
4
5
6
temp1 <- sample(1:200, 150)
temp2 <- list()
for(i in temp1) temp2 <- appendToList(temp2, imgmods[[i]])
temp3 <- imglabels[temp1]
# de-activated because this process is very long...
#temp4 <- sampleClassif(temp2, temp3)

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.