Abstract: Training deep/convolutional neural networks (DNNs/CNNs) requires a large amount of memory and iterative computation, which necessitates speedup and energy reduction, especially for edge ...