% GMMB_CMEANS  simple c-means clustering
 %
 % T = CMEANS(data, nclust, count)
 % [T, CLUST] = CMEANS(...)
 %
 % data    input data, N x D matrix
 % nclust  number of clusters
 % count   number of iterations
 %
 % T       output, data labels, 1 x N vector
 % CL      output, cluster centers, nclust x D matrix
 %
 % Author: Jarmo Ilonen
 % Editor: Pekka Paalanen
 %
 % $Name:  $ $Id: gmmb_cmeans.m,v 1.1 2004/08/16 15:06:44 paalanen Exp $

 function [pclass, clust]=gmmb_cmeans(pdata,nclust,count)

 rp = randperm(size(pdata,1));
 clust = pdata(rp(1:nclust),:);
 
for kierros=1:count
    % lasketaan et?isyyden neli? jokaisesta samplesta jokaisen klusterin
     % keskipisteeseen
     for i=1:nclust
         vd = pdata - repmat(clust(i,:),size(pdata,1),1);
        cet(:,i) = sum(abs(vd).^2, 2);
    end
 
    % lasketaan klustereille uudet keskipisteet
    [a, pclass]=min(cet');

     for i=1:nclust
        clust(i,:) = mean( pdata(find(pclass==i), :) );
    end
 end