HAWQ: Hessian AWare Quantization of Neural Networks With Mixed-Precision | IEEE Conference Publication | IEEE Xplore