Back to Search Start Over

NG+ : A Multi-Step Matrix-Product Natural Gradient Method for Deep Learning

Authors :
Yang, Minghan
Xu, Dong
Cui, Qiwen
Wen, Zaiwen
Xu, Pengxiang
Yang, Minghan
Xu, Dong
Cui, Qiwen
Wen, Zaiwen
Xu, Pengxiang
Publication Year :
2021

Abstract

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 fisher information matrix (GFIM) using the products of gradients in the matrix form rather than the traditional vectorization. Then, our generalized natural gradient direction is simply the inverse of the GFIM multiplies the gradient in the matrix form. Moreover, the GFIM and its inverse keeps the same for multiple steps so that the computational cost can be controlled and is comparable with the first-order methods. A global convergence is established under some mild conditions and a regret bound is also given for the online learning setting. Numerical results on image classification with ResNet50, quantum chemistry modeling with Schnet, neural machine translation with Transformer and recommendation system with DLRM illustrate that GN+ is competitive with the state-of-the-art methods.

Details

Database :
OAIster
Publication Type :
Electronic Resource
Accession number :
edsoai.on1269557574
Document Type :
Electronic Resource