Deep Learning, Part 3: Too Deep or Not Too Deep? That is the Question