Addition is almost all you need: Compressing neural networks with double binary factorization

MQRe-assign community