Scaling Neural Network Performance through Customized Hardware Architectures on Reconfigurable Logic
2018·Arxiv