1 citations · 1 across the 1 of their papers we have counts for
3 papers
math.OC2021★ 1 cited
NG+ : A Multi-Step Matrix-Product Natural Gradient Method for Deep Learning
Minghan Yang, Dong Xu, Qiwen Cui +2
In this paper, a novel second-order method called NG+ is proposed. By following the rule ``the shape of the gradient equals the shape of the parameter", we define a generalized fis…
math.OC2020
Enhance Curvature Information by Structured Stochastic Quasi-Newton Methods
Minghan Yang, Dong Xu, Hongyu Chen +2
In this paper, we consider stochastic second-order methods for minimizing a finite summation of nonconvex functions. One important key is to find an ingenious but cheap scheme to i…
math.OC2020
Sketchy Empirical Natural Gradient Methods for Deep Learning
Minghan Yang, Dong Xu, Zaiwen Wen +2
In this paper, we develop an efficient sketchy empirical natural gradient method (SENG) for large-scale deep learning problems. The empirical Fisher information matrix is usually l…