Some Experiments Using Extra Output Learning to Hint Multi Layer Perceptrons