Abstract: Deep neural network (DNN) training is an iterative process of updating network weights, called gradient computation, where (mini-batch) stochastic gradient descent (SGD) algorithm is ...