Chaochao Lin

Chaochao Lin

Auto-generated profile, based on publicly available information

Chaochao Lin

School of Computer Science, Beijing Institute of Technology, Beijing, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 January 2024 in Computer Methods and Programs in Biomedicine
Verified
Journal Article
Closed access
WLWeihua Liu
Weihua Liu
CLChaochao Lin
Chaochao Lin
DCDuanduan Chen
Duanduan Chen
LNLijuan Niu
Lijuan Niu
RZRui Zhang
Rui Zhang
ZPZhaoqiong Pi
Zhaoqiong Pi

11 March 2026 in Reading and Writing
Verified
Journal Article
Closed access
PDF file
WZWei Zhou
Wei Zhou
RWRan Wang
Ran Wang
YXYan Xiong
Yan Xiong
XZXinshuang Zhang
Xinshuang Zhang
XLXiabi Liu
Xiabi Liu
QSQi Shu
Qi Shu

10 December 2024 in arXiv
Verified
Preprint
Open access
PDF file
WLWeihua Liu
Weihua Liu
said boumarafsaid boumaraf
said boumaraf
JLJianwu Li
Jianwu Li
CLChaochao Lin
Chaochao Lin
XLXiabi Liu
Xiabi Liu
NWNaoufel Werghi
Naoufel Werghi
Natural gradient descent (NGD) is a powerful optimization technique for machine learning, but the computational complexity of the inverse Fisher information matrix limits its application in training deep neural networks. To overcome this challenge, we propose a novel optimization method for training deep neural networks called structured natural gradient descent (SNGD). Theoretically, we demonstrate that optimizing the original network using NGD is equivalent to using fast gradient descent (GD) to optimize the reconstructed network with a structural transformation of the parameter matrix. Thereby, we decompose the calculation of the global Fisher information matrix into the efficient computation of local Fisher matrices via constructing local Fisher layers in the reconstructed network to speed up the training. Experimental results on various deep networks and datasets demonstrate that SNGD achieves faster convergence speed than NGD while retaining comparable solutions. Furthermore, our method outperforms traditional GDs in terms of efficiency and effectiveness. Thus, our proposed method has the potential to significantly improve the scalability and efficiency of NGD in deep learning applications. Our source code is available at https://github.com/Chaochao-Lin/SNGD.

24 August 2024 in Pattern Recognition Letters
Verified
Journal Article
Closed access
WLWeihua Liu
Weihua Liu
XLXiabi Liu
Xiabi Liu
HLHuiyu Li
Huiyu Li
CLChaochao Lin
Chaochao Lin