1 paper
Dong Wang, Yicheng Liu, Wenwo Tang +4
In this paper, we propose a new first-order gradient-based algorithm to train deep neural networks. We first introduce the sign operation of stochastic gradients (as in sign-based…