EUSPACE Convert PE space to Euclidean space
B = A*EUSPACE
DescriptionThe PseudoEuclidean (PE) space consist of two Euclidean subspaces, the positive one and the negative one. A pseudoEuclidean mapping as constructed by PE_EM generates vectors for which the first p components refer to the positive space and the next q components to the negative space. The signature (p,q) can be retrieved by GETSIG. This routine returns for TYPE = 'ass' the socalled associative space. B contains thereby the same set of vectors, but now with signature (p+q,0). For TYPE = 'pos' just the positive part is returned, B = A(:,1:p). The signature is now just (p,0). For TYPE = 'neg', the negative part is returned, B = A(:,p+1:p+q) and the signature is set to (q,0) by which it is interpreted as Euclidean. These conversions influence various mappings and classifiers defined for the PE space. TYPE = 'dec' determines the Euclidean correction found by adding a sufficiently large positive constant to the offdiagonal elements of the squared original dissmilarity matrix. Reference(s)[1] R.P.W. Duin, E. Pekalska, A. Harol, W.J. Lee, and H. Bunke, On Euclidean corrections for nonEuclidean dissimilarities, in: Proc. SSPR & SPR 2008, LNCS 5342, Springer, 2008, 551561. [2] R.P.W. Duin, E. Pekalska, NonEuclidean Dissimilarities: Causes and Informativeness, in: SSPR & SPR 2010, LNCS 6218, Springer, 2010, 324333. See alsodatasets, pe_em, getsig, pe_distm, pe_knnc, pe_parzenc, pe_nmc, pe_kernelm, pe_affine, pe_mtimes,
