Scaling Neural Network Performance through Customized Hardware Architectures on Reconfigurable Logic | IEEE Conference Publication | IEEE Xplore