中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 深海科学与工程研究所
  3. 中国科学院深海科学与工程研究所
  4. 深海工程技术部
  5. 深海信息技术研究室
Learning dynamics of kernel-based deep neural networks in manifolds

文献类型：期刊论文

作者Wu, Wei2,4; Jing, Xiaoyuan2,3; Du, Wencai5; Chen, Guoliang1
刊名SCIENCE CHINA-INFORMATION SCIENCES
出版日期2021-11-01
卷号64期号:11页码:15
关键词learning dynamics kernel-based convolution manifolds control model network stability
ISSN号1674-733X
DOI10.1007/s11432-020-3022-3
通讯作者Jing, Xiaoyuan
英文摘要Convolutional neural networks (CNNs) obtain promising results via layered kernel convolution and pooling operations, yet the learning dynamics of the kernel remain obscure. We propose a continuous form to describe kernel-based convolutions through integration in neural manifolds. The status of spatial expression is proposed to analyze the stability of kernel-based CNNs. We divide CNN dynamics into the three stages of unstable vibration, collaborative adjusting, and stabilized fluctuation. According to the system control matrix of the kernel, the kernel-based CNN training proceeds via the unstable and stable status and is verified by numerical experiments.
WOS关键词SINGULARITIES ; WORKS
资助项目Key Project of National Natural Science Foundation of China[61933013] ; Strategic Priority Research Program of the Chinese Academy of Sciences[XDA22030301] ; NSFC-Key Project of General Technology Fundamental Research United Fund[U1736211] ; Natural Science Foundation of Guangdong Province[2019A1515011076] ; Key Project of Natural Science Foundation of Hubei Province[2018CFA024]
WOS研究方向Computer Science ; Engineering
语种英语
WOS记录号WOS:000709794800002
出版者SCIENCE PRESS
资助机构Key Project of National Natural Science Foundation of China ; Strategic Priority Research Program of the Chinese Academy of Sciences ; NSFC-Key Project of General Technology Fundamental Research United Fund ; Natural Science Foundation of Guangdong Province ; Key Project of Natural Science Foundation of Hubei Province
源URL[http://ir.idsse.ac.cn/handle/183446/9420]  
专题深海工程技术部_深海信息技术研究室
通讯作者Jing, Xiaoyuan
作者单位1.Shenzhen Univ, Coll Comp Sci & Software Engn, Shenzhen 518060, Peoples R China
2.Wuhan Univ, Sch Comp Sci, Wuhan 430072, Peoples R China
3.Guangdong Univ Petrochem Technol, Sch Comp, Maoming 525000, Peoples R China
4.Chinese Acad Sci, Inst Deep Sea Sci & Engn, Sanya 572000, Peoples R China
5.City Univ Macau, Inst Data Sci, Macau 999078, Peoples R China
推荐引用方式
GB/T 7714		 Wu, Wei,Jing, Xiaoyuan,Du, Wencai,et al. Learning dynamics of kernel-based deep neural networks in manifolds[J]. SCIENCE CHINA-INFORMATION SCIENCES,2021,64(11):15.
APA Wu, Wei,Jing, Xiaoyuan,Du, Wencai,&Chen, Guoliang.(2021).Learning dynamics of kernel-based deep neural networks in manifolds.SCIENCE CHINA-INFORMATION SCIENCES,64(11),15.
MLA Wu, Wei,et al."Learning dynamics of kernel-based deep neural networks in manifolds".SCIENCE CHINA-INFORMATION SCIENCES 64.11(2021):15.

入库方式： OAI收割

来源：深海科学与工程研究所

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。