KMeans in Octave

gistfile1.txt

function [cent, idx, sizes, distortion]=km(data, k, maxIter)

N = size(data,1);
COL = size(data,2);

cent = data(randperm(N)(1:k), : );
#cent = rand(k, COL) * 0.5 + 0.5
for iter = 1 : maxIter
	fprintf("Iteration %d\n", iter);
	dist = [];
	for i = 1 : k
		dist = [dist ; sumsq( (data - repmat(cent(i,:),N, 1))')];
	endfor

	[val, idx] = min(dist);
	cent = [];
	for i = 1: k
		cent = [cent; mean(data(idx == i,:) ,1)];
	endfor

	sizes = [];
	for i = 1 : k
        	sizes = [sizes; length(idx(idx == i))];
	endfor
	sizes'
endfor

distortion = 0;
for i = 1 : k
	distortion = distortion + sum(sumsq((data(idx == i, :) - repmat(cent(i,:), sizes(i), 1)) )) / sizes(i);
endfor

distortion