% computes the NC centroids corresponding to the given points using K-Means
function centroids = clustering_pc(points, NC)
  [n m] = size(points);
  for i = 1:NC
    if i > 1
      y = randi(n);
      if x == y
        y = randi(n);
      endif
      x = y;
    else
      x = randi(n);
    endif
    centroids(i,:) = points(x,:);
  endfor

  prev_centroids = zeros(NC, m);
  error = 0.0001;
  while (abs(prev_centroids - centroids) > error)
    poz = zeros(n,1);
    prev_centroids = centroids;
    for i = 1:n
      k = 1;
      dist_min = norm(points(i,:) - centroids(1,:));
      for j = 2:NC
        dist = norm(points(i,:) - centroids(j,:));
        if (dist < dist_min)
          dist_min = dist;
          k = j;
        endif
      endfor
      poz(i) = k;
    endfor
    val = 1;
    while (val <= NC)
      nr_points = 0;
      sumX = sumY = sumZ = 0;
      for m = 1:n
        if (val == poz(m))
          sumX = sumX + points(m, 1);
          sumY = sumY + points(m, 2);
          sumZ = sumZ + points(m, 3);
          nr_points = nr_points + 1;
        endif
      endfor
      centroids(val, :) = [sumX sumY sumZ] * (1/nr_points);
      val = val + 1;
    endwhile
  endwhile

    % TODO K-Means code
end