Gradient Based Optimization in Deep Learning