Simplifying Deep Learning and Neural Networks for Embedded Processing