source: distools/disrc.m @ 13

%DISRC Dissimilarity based classifier by distance ranking
%
%       W = disrc(A)
%       W = A*disrc
%
% A should be a dissimilarity based dataset with class labels as
% feature labels. The classifier W is based on distance ranking.
% A new dataset B defined by dissimilarities to the same representation
% set A is classified by D = B*W by comparing the dissimilarities in B
% with the ranked dissimilarities in A for each object in the
% representation set separately. The columns in D yield for each
% object in the representation set a distance such that D*classc
% generates for these objects the ranked based probabilities.
%
% Formally W is a set of parallel mappings, so B*W*classd combines
% them by assigning objects to the class with the highest probability.
% Other combiners may be used as well, e.g. B*W*classc*prodc,
% combining all probabilities by the product rule.
%
% See mappings, classc, classd.
%
% $Id: disrc.m,v 1.2 2001/07/10 16:35:58 pavel Exp $

% Copyright: R.P.W. Duin, duin@ph.tn.tudelft.nl
% Faculty of Applied Physics, Delft University of Technology
% P.O. Box 5046, 2600 GA Delft, The Netherlands

function w = disrc(a,v)

if nargin < 1 | isempty(a) % empty call (untrained classifier)
        w = mapping('disrc');
                
elseif nargin == 1    % training
        [nlab,lablist,m,k,c,p] = dataset(a);
        [nn,nf,fl] = renumlab(lablist,getfeat(a));
        if c < 2
                error('Dataset should containf more than one class')
        end
        if max(nf) > c
                error('Feature labels of dataset do not match with class labels')
        end
        w = [];
        for j = 1:k
                b = {+a(find(nlab==nf(j)),j) +a(find(nlab~=nf(j)),j)};
                w = [w; mapping('disrc',b,fl(nf(j),:),1,1,1)];
        end

%       w = cnormc(w,a); forget normalisation and apriori probs for the moment

elseif nargin == 2      % evaluation
        [m,k] = size(a);
        b = +v; b1 = b{1}; b2 = b{2};
        n1 = length(b1); n2 = length(b2);
        w1 = zeros(m,1); w2 = zeros(m,1);
        [d R1] = sort([b1;+a]);
        [d R2] = sort([b2;+a]);
        L1 = find(R1>n1);
        L2 = find(R2>n2);
        w1(R1(L1)-n1) = (n1+2-(L1-[1:m]'))/(n1+2);
        w2(R2(L2)-n2) = (n2+2-(L2-[1:m]'))/(n2+2);
%       disp([w1;w2])
        w = invsig(dataset((w1+w2)/2,getlab(a),getfeat(v)));
else
        error('Illegal call')
end