DISPN

Split dissimilarity matrix in positive and negative part

    [DP,DN,W] = DISPOSNEG(D)

Input
D	Full, square dissimilarity matrix, dataset or double

Output
DP	Euclidean distance matrix of the positive PE space
DN	Euclidean distance matrix of the negative PE space
W	Mapping to PE space

Description

D is made square (by (D+D')/2) and its diagonal is set to zero. Next a  pseudo-Euclidean embedding (PE) is made. The positive and negative  dimensions are separated. In the two resulting space the Euclidean  distances of the data are found. The following relations hold
D.^2 = DP.^2 - DN.^2; NEF(DP) = 0, NEF(DN) = 0; X = D*W;