Understanding Neural Network Binarization with Forward and Backward Proximal Quantizers | Read Paper on Bytez