1 code implementation • 3 Dec 2020 • Jinhuan Duan, Xianxian Li, Shiqi Gao, Jinyan Wang, Zili Zhong
In this paper, to prevent gradient leakage while keeping the accuracy of model, we propose the super stochastic gradient descent approach to update parameters by concealing the modulus length of gradient vectors and converting it or them into a unit vector.
