Why Deep Learning (DL) algorithms need at least double or several folds of data for better pattern analysis (or best pattern detector algorithms) in comparison to the conventional three layers or single step algorithms/methods for pattern analysis?????