An Energy-Efficient Deep Convolutional Neural Network Training Accelerator for I

Abstract

A scalable deep-learning accelerator supporting the training process is implemented for device personalization of deep convolutional neural networks (CNNs). It consists of three proces- sor cores operating with distinct energy-efficient dataflow for dif- ferent types of computation in CNN training. Unlike the previous works where they implement design techniques to exploit the same characteristics from the inference, we analyze major issues that occurred from training in a resource-constrained sys