Hardware-Friendly Acceleration for Deep Neural Networks with Micro-Structured Compression | IEEE Conference Publication | IEEE Xplore