Loss And Loss Functions For Training Deep Learning Neural Networks