Compressing Low Precision Deep Neural Networks Using Sparsity-Induced Regularization in Ternary Networks | Read Paper on Bytez