salso (version 0.1.9)

binder: Compute a Partition Loss Function

Description

These functions compute the expectation of the Binder loss and the lower bound of the expectation of the variation of information loss for given partitions based on the supplied pairwise similarity matrix.

Usage

binder(partitions, psm)
VI.lb(partitions, psm)

Arguments

partitions

An integer matrix of cluster labels, where each row is a partition given as cluster labels. Two items are in the same subset (i.e., cluster) if their labels are equal.

psm

A pairwise similarity matrix, i.e., n-by-n symmetric matrix whose (i,j) element gives the (estimated) probability that items i and j are in the same subset (i.e., cluster) of a partition (i.e., clustering).

Value

A numeric vector of length equal to the number of rows of partitions, where each element gives the value of the loss function.

Examples

# NOT RUN {
probs <- psm(iris.clusterings, parallel=FALSE)
binder(iris.clusterings[1:5,], probs)
VI.lb(iris.clusterings[1:5,], probs)

# }