Understanding Deep Learning Through Empirical Falsification