Towards understanding residual and dilated dense neural networks via convolutional sparse coding

doi:10.1093/nsr/nwaa159

CSpace > 应用数学研究所

	Towards understanding residual and dilated dense neural networks via convolutional sparse coding
	Zhang, Zhiyang 1,2; Zhang, Shihua1,2,3,4
	2021-03-01
发表期刊	NATIONAL SCIENCE REVIEW
ISSN	2095-5138
卷号	8 期号:3 页码:13
摘要	Convolutional neural network (CNN) and its variants have led to many state-of-the-art results in various fields. However, a clear theoretical understanding of such networks is still lacking. Recently, a multilayer convolutional sparse coding (ML-CSC) model has been proposed and proved to equal such simply stacked networks (plain networks). Here, we consider the initialization, the dictionary design and the number of iterations to be factors in each layer that greatly affect the performance of the ML-CSC model. Inspired by these considerations, we propose two novel multilayer models: the residual convolutional sparse coding (Res-CSC) model and the mixed-scale dense convolutional sparse coding (MSD-CSC) model. They are closely related to the residual neural network (ResNet) and the mixed-scale (dilated) dense neural network (MSDNet), respectively. Mathematically, we derive the skip connection in the ResNet as a special case of a new forward propagation rule for the ML-CSC model. We also find a theoretical interpretation of dilated convolution and dense connection in the MSDNet by analyzing the MSD-CSC model, which gives a clear mathematical understanding of each. We implement the iterative soft thresholding algorithm and its fast version to solve the Res-CSC and MSD-CSC models. The unfolding operation can be employed for further improvement. Finally, extensive numerical experiments and comparison with competing methods demonstrate their effectiveness.
关键词	convolutional neural network convolutional sparse coding residual neural network mixed-scale dense neural network dilated convolution dense connection
DOI	10.1093/nsr/nwaa159
收录类别	SCI
语种	英语
资助项目	National Key Research and Development Program of China[2019YFA0709501] ; National Natural Science Foundation of China[11661141019] ; National Natural Science Foundation of China[61621003] ; National Ten Thousand Talent Program for Young Top-notch Talents ; CAS Frontier Science Research Key Project for Top Young Scientists[QYZDB-SSW-SYS008]
WOS研究方向	Science & Technology - Other Topics
WOS类目	Multidisciplinary Sciences
WOS记录号	WOS:000649002700009
出版者	OXFORD UNIV PRESS
引用统计
文献类型	期刊论文
条目标识符	http://ir.amss.ac.cn/handle/2S8OKBNM/58732
专题	应用数学研究所
通讯作者	Zhang, Shihua
作者单位	1.Chinese Acad Sci, Acad Math & Syst Sci, Natl Ctr Math & Interdisciplinary Sci, Ctr Excellence Math Sci,Key Lab Random Complex St, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Sch Math Sci, Beijing 100049, Peoples R China 3.Chinese Acad Sci, Ctr Excellence Anim Evolut & Genet, Kunming 650223, Yunnan, Peoples R China 4.Univ Chinese Acad Sci, Chinese Acad Sci, Hangzhou Inst Adv Study, Key Lab Syst Biol, Hangzhou 310024, Peoples R China
推荐引用方式 GB/T 7714	Zhang, Zhiyang,Zhang, Shihua. Towards understanding residual and dilated dense neural networks via convolutional sparse coding[J]. NATIONAL SCIENCE REVIEW,2021,8(3):13.
APA	Zhang, Zhiyang,&Zhang, Shihua.(2021).Towards understanding residual and dilated dense neural networks via convolutional sparse coding.NATIONAL SCIENCE REVIEW,8(3),13.
MLA	Zhang, Zhiyang,et al."Towards understanding residual and dilated dense neural networks via convolutional sparse coding".NATIONAL SCIENCE REVIEW 8.3(2021):13.