clf_knn_dist

 k-nearest neighbor classifier based on a distance matrix D.

 k==1 is much faster than k>1.  For k>1, ties are broken randomly.

 INPUTS
   D       - MxN array of distances from M-TEST points to N-TRAIN points.
   IDX     - ntrain length vector of class memberships 
             if IDX(i)==IDX(j) than sample i and j are part of the same class
   k       - [optional] number of nearest neighbors to use, 1 by default

 OUTPUTS
   IDXpred - length M vector of classes for training data

 EXAMPLE
   % [given D and IDX]
   for k=1:size(D,2) err(k)=sum(IDX==clf_knn_dist(D,IDX,k)); end; 
   figure(1); plot(err)

 DATESTAMP
   11-Oct-2005  8:00pm

 See also CLF_KNN