RecFac

Location: ..\cubatch\Common\RecFac.m

function [Recovery,Perm] = RecFac(Meas,a,b,c,...,A,B,C,...)

Description:
If Meas is 'MSE' the factors have to be previously ordered and gives the mean square error between the first set of vectors and the second: the m-th element of the n-th row is the error for the m-th vector in the n-th dimension (or loading matrix).
Perm is in this case an identity matrix as this method does not seek the correct order of the factors.

If Meas is 'MitBur' the function checks all the possible permutations of factors in the first set opting for the one having the maximum average congruence with the second set. Perm, is the corresponding permutation matrix. The recovery is given per factor and is the cosine (congruence) between the component of the first set and the corresponding ones in the second.

The number of factors in the two sets can be different, in the latter case the set with the highest number of factors must come first. In such a case the permutation matrix will be F x F, where F is the number of factors in the first set. The resulting permutation will cause the exceeding factors in the "permuted" matrices to be columns of zeros.

Inputs:
Meas: type of quality of the recovery requested it can either be 'MitBur' or 'MSE'
A,B,C,...: set of factors 1
a,b,c,...: set of factors 2

Outputs:
Recovery: quality of the recovery
Perm: permutation matrix

Called by:
ValidateModel, Model_PARAFAC\Compute_Risk

Subroutines:
Internal: None
External: NWay\NormIt

Author:
Giorgio Tomasi
Royal Agricultural and Veterinary University
MLI, LMT, Chemometrics group
Rolighedsvej 30
DK-1958 Frederiksberg C
Danmark

Last modified: 21-Oct-2002 08:22:16

Contact: Giorgio Tomasi, gt@kvl.dk

References