TRAIN_SIZE = 0.75;
objects = [10, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000];
resize = [8, 16, 32];
pca = [0.80, 0.85, 0.90, 0.95, 0.99, 100];
t = [];
d = [];
global pixels;

min_test = 1;
min_benchmark = 1;

for i=1:size(objects, 2)
  per_class = objects(i);
  a = prnist([0:9],[1:per_class]);
  for j=1:size(resize, 2)
    pixels = resize(j);
    resampled = my_rep(a);
    [train, tst] = gendat(resampled, TRAIN_SIZE)
    for k=1:size(pca, 2)
      cumulative = pca(k);
      if cumulative < 100
        pca_mapping = pcam(cumulative);
      else
        pca_mapping = 1;
      end

      tic;
      classifier = pca_mapping * my_prox('d', 1);
      classifier = classifier(train);
      train_time = toc;
      tic;
      test_err = tst * classifier * testc;
      test_time = toc;

      if test_err < min_test
        min_test = test_err;
        min_pixels = pixels;
        min_pca = cumulative;
        min_objects = per_class;
      end

      benchmark = nist_eval('my_rep', classifier);
      min_benchmark = min([benchmark min_benchmark]);
    end
  end
end

min_test = test_err
min_pixels = pixels
min_pca = cumulative
min_objects = per_class