Weight-Oriented Approximation for Energy-Efficient Neural Network Inference Accelerators | IEEE Journals & Magazine | IEEE Xplore