Deep learning products tend to raise their precision with the escalating total of coaching data, While classic equipment learning products like SVM and naive Bayes classifier cease increasing after a saturation level. This paper showed that supervised coaching of very deep neural networks is