Article(id=1329455739829072004, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329455653363495339, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202220483, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1654790400000, receivedDateStr=2022-06-10, revisedDate=1681747200000, revisedDateStr=2023-04-18, acceptedDate=null, acceptedDateStr=null, onlineDate=1794896548467, onlineDateStr=2026-11-17, pubDate=1706803200000, pubDateStr=2024-02-02, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794896548467, onlineIssueDateStr=2026-11-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794896548467, creator=13701087609, updateTime=1794896548467, updator=13701087609, issue=Issue{id=1329455653363495339, tenantId=1146029695717560320, journalId=1324759914146529478, year='2024', volume='61', issue='2', pageStart='273', pageEnd='538', issueExtLink='null', onlineDate='null', pubDate='1706803200000', pubDateStr='2024-02-02', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794896527853, creator='13701087609', updateTime=1794896527853, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext=null, issueFiles=null, downloadFileDto=null}, startPage=362, endPage=384, ext={EN=ArticleExt(id=1329455740030398597, articleId=1329455739829072004, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Survey on Hypergraph Neural Networks, columnId=1329455694815802127, journalTitle=Journal of Computer Research and Development, columnName=Artificial Intelligence and Pattern Recognition, runingTitle=null, highlight=null, articleAbstract=

In recent years, graph neural networks have achieved remarkable results in application fields such as recommendation systems and natural language processing with the help of large amounts of data and supercomputing power, and they mainly deal with graph data with pairwise relationships. However, in many real-world networks, the relationships between objects are more complex and beyond pairwise, such as scientific collaboration networks, protein networks, and others. If we directly use a graph to represent the complex relationships as pairwise relations, which will lead to a loss of information. Hypergraph is a flexible modeling tool, which shows higher-order relationships that cannot be fully described by a graph, making up for the shortage of graph. In light of this, scholars begin to care about how to develop neural networks on hypergraph, and successively put forward many hypergraph neural network models. Therefore, we overview the existing hypergraph neural network models. Firstly, we comprehensively review the development of the hypergraph neural network in the past three years. Secondly, we propose a new classification method according to the design method of hypergraph neural networks, and elaborate on representative models. Then, we introduce the application areas of hypergraph neural networks. Finally, the future research direction of hypergraph neural networks are summarized and discussed.

, authors=Jingjing Lin1, 2, 3, Zhonglin Ye1, 2, Haixing Zhao1, 2, *, Zhuoran Li1, 2, authorsList=Jingjing Lin, Zhonglin Ye, Haixing Zhao, Zhuoran Li, authorCompany=null, correspAuthors=Haixing Zhao, authorNote=

bron in 1986. PhD candidate, lecturer. Her main research interests include graph neural networks and hypergraph neural networks

born in 1989. PhD, associate professorr, PhD supervisor. Member of CCF. His main research interests include graph neural networks, knowledge extraction, and network representation learning

born in 1969. PhD, professor, PhD supervisor. Member of CCF. His main research interests include complex network, graph neural networks, machine translation, hypergraph theory, and network reliability

born in 1996. Master. His main research interests include data mining and graph neural networks

, correspAuthorsNote=null, copyrightStatement=Copyright © 2024 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1329455742928662678, articleId=1329455739829072004, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=超图神经网络综述, columnId=1329455694975185681, journalTitle=计算机研究与发展, columnName=人工智能和模式识别, runingTitle=null, highlight=null, articleAbstract=

近年来,图神经网络借助大量数据和超强计算能力在推荐系统和自然语言处理等应用领域取得显著成效,它主要处理具有成对关系的图数据. 但许多现实网络中的对象之间的关系是复杂的非成对关系,如科研合作网络、蛋白质网络等. 若直接用图结构将这种复杂的关系表示为成对关系,会导致信息丢失. 超图是一种灵活的建模工具,可以展现出图无法完整刻画的高阶关系,弥补了图的不足. 鉴于此,研究者开始关心如何在超图上设计神经网络,并相继提出应用于下游任务的超图神经网络模型(hypergraph neural network,HGNNs). 故对现有的超图神经网络模型进行综述,首先全面回顾超图神经网络在过去3年的研究历程;其次根据设计超图神经网络采用的方法不同对其进行分类,并详细地阐述代表性的模型;然后介绍了超图神经网络的应用领域;最后总结和探讨了超图神经网络未来的研究方向.

, authors=林晶晶1, 2, 3, 冶忠林1, 2, 赵海兴1, 2, *, 李卓然1, 2, authorsList=林晶晶, 冶忠林, 赵海兴, 李卓然, authorCompany=null, correspAuthors=赵海兴, authorNote=

1986年生. 博士研究生,讲师. 主要研究方向为图神经网络、超图神经网络

1989年生. 博士,副教授,博士生导师. CCF会员. 主要研究方向为图神经网络、知识抽取、网络表示学习

1969年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为复杂网络、图神经网络、机器翻译、超图理论、网络可靠性

1996年生. 硕士. 主要研究方向为数据挖掘、图神经网络

, correspAuthorsNote=
赵海兴()
, copyrightStatement=版权所有 © 2024 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=20u8lOWzt4ZzMKXXf7Zfeg==, magXml=DE1MUvhqPKLrzJTTuwzPVg==, pdfUrl=null, pdf=fNA1eeyJGP/eTDzIhS+Tzg==, pdfFileSize=1402750, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=zJaNaA/TKhRzd+xGCe4HfA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=/sbHaR8bQ4yan+VDFcPjnA==, mapNumber=null, fund=null)}, authors=[Author(id=1329455743402619041, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=ljj_mail@126.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329455743478116517, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743402619041, language=EN, stringName=Jingjing Lin, firstName=Jingjing, middleName=null, lastName=Lin, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 3, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008
3Department of Information Engineering, Xining Urban Vocational & Technical College, Xining 810003, bio={"content":"

bron in 1986. PhD candidate, lecturer. Her main research interests include graph neural networks and hypergraph neural networks

"}, bioImg=null, bioContent=

bron in 1986. PhD candidate, lecturer. Her main research interests include graph neural networks and hypergraph neural networks

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455743830438054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743402619041, language=CN, stringName=林晶晶, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 3, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008
3西宁城市职业技术学院信息工程系 西宁 810003, bio={"img":"G5rQbtrDWA86ZLRSo9TXJg==","content":"

1986年生. 博士研究生,讲师. 主要研究方向为图神经网络、超图神经网络

"}, bioImg=G5rQbtrDWA86ZLRSo9TXJg==, bioContent=

1986年生. 博士研究生,讲师. 主要研究方向为图神经网络、超图神经网络

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)]), AuthorCompany(id=1329455743322927261, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=3, ext=[AuthorCompanyExt(id=1329455743327121566, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743322927261, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Department of Information Engineering, Xining Urban Vocational & Technical College, Xining 810003), AuthorCompanyExt(id=1329455743335510175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743322927261, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3西宁城市职业技术学院信息工程系 西宁 810003)])]), Author(id=1329455743893352616, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329455743989821611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743893352616, language=EN, stringName=Zhonglin Ye, firstName=Zhonglin, middleName=null, lastName=Ye, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008, bio={"content":"

born in 1989. PhD, associate professorr, PhD supervisor. Member of CCF. His main research interests include graph neural networks, knowledge extraction, and network representation learning

"}, bioImg=null, bioContent=

born in 1989. PhD, associate professorr, PhD supervisor. Member of CCF. His main research interests include graph neural networks, knowledge extraction, and network representation learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455744086290604, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743893352616, language=CN, stringName=冶忠林, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008, bio={"img":"VVg0HWqw4/9zUsqlvaF/EA==","content":"

1989年生. 博士,副教授,博士生导师. CCF会员. 主要研究方向为图神经网络、知识抽取、网络表示学习

"}, bioImg=VVg0HWqw4/9zUsqlvaF/EA==, bioContent=

1989年生. 博士,副教授,博士生导师. CCF会员. 主要研究方向为图神经网络、知识抽取、网络表示学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)])]), Author(id=1329455744165982382, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=h.x.zhao@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329455744241479857, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744165982382, language=EN, stringName=Haixing Zhao, firstName=Haixing, middleName=null, lastName=Zhao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008, bio={"content":"

born in 1969. PhD, professor, PhD supervisor. Member of CCF. His main research interests include complex network, graph neural networks, machine translation, hypergraph theory, and network reliability

"}, bioImg=null, bioContent=

born in 1969. PhD, professor, PhD supervisor. Member of CCF. His main research interests include complex network, graph neural networks, machine translation, hypergraph theory, and network reliability

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455744312783026, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744165982382, language=CN, stringName=赵海兴, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008, bio={"img":"pS/TFCPeGGYENBVzrfBeOA==","content":"

1969年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为复杂网络、图神经网络、机器翻译、超图理论、网络可靠性

"}, bioImg=pS/TFCPeGGYENBVzrfBeOA==, bioContent=

1969年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为复杂网络、图神经网络、机器翻译、超图理论、网络可靠性

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)])]), Author(id=1329455744400863412, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329455744480555191, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744400863412, language=EN, stringName=Zhuoran Li, firstName=Zhuoran, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008, bio={"content":"

born in 1996. Master. His main research interests include data mining and graph neural networks

"}, bioImg=null, bioContent=

born in 1996. Master. His main research interests include data mining and graph neural networks

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455744543469752, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744400863412, language=CN, stringName=李卓然, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008, bio={"img":"cMiNtBIHDxwc9c2wbVX1Bg==","content":"

1996年生. 硕士. 主要研究方向为数据挖掘、图神经网络

"}, bioImg=cMiNtBIHDxwc9c2wbVX1Bg==, bioContent=

1996年生. 硕士. 主要研究方向为数据挖掘、图神经网络

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)])])], keywords=[Keyword(id=1329455746200219834, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, orderNo=1, keyword=hypergraph), Keyword(id=1329455746267328699, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, orderNo=2, keyword=hypergraph neural network), Keyword(id=1329455746330243260, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, orderNo=3, keyword=classification), Keyword(id=1329455746397352125, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, orderNo=4, keyword=graph), Keyword(id=1329455746502209726, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, orderNo=5, keyword=graph neural networks), Keyword(id=1329455746577707199, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, orderNo=1, keyword=超图), Keyword(id=1329455746628038848, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, orderNo=2, keyword=超图神经网络), Keyword(id=1329455746690953409, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, orderNo=3, keyword=分类), Keyword(id=1329455746762256578, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, orderNo=4, keyword=图), Keyword(id=1329455746833559747, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, orderNo=5, keyword=图神经网络)], refs=[Reference(id=1329455752437149937, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2012, 60(6): 84−90, articleTitle=null, refAbstract=null), Reference(id=1329455752508453106, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Elman J L. Distributed representations, simple recurrent networks, and grammatical structure[J]. Machine Learning, 1991, 7(2): 195−225., articleTitle=null, refAbstract=null), Reference(id=1329455752596533491, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436−444, articleTitle=null, refAbstract=null), Reference(id=1329455752655253748, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735−1780, articleTitle=null, refAbstract=null), Reference(id=1329455752797860085, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Chung J, Gülçehre Ç, Cho K, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[J]. arXiv preprint, arXiv: 1412. 3555, 2014, articleTitle=null, refAbstract=null), Reference(id=1329455752860774646, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=徐冰冰,岑科廷,黄俊杰,等. 图卷积神经网络综述[J]. 计算机学报,2020,43(5):755−780, articleTitle=null, refAbstract=null), Reference(id=1329455752969826551, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Bingbing, Cen Keting, Huang Junjie, et al. A survey on graph convolutional neural network[J]. Chinese Journal of Computers, 2020, 43(5): 755−780(in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329455753057906936, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Bruna J, Zaremba W, Szlam A, et al. Spectral networks and locally connected networks on graphs[J]. arXiv preprint, arXiv: 1312. 6203, 2014, articleTitle=null, refAbstract=null), Reference(id=1329455753125015801, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=马帅,刘建伟,左信. 图神经网络综述[J]. 计算机研究与发展,2022,59(1):47−80, articleTitle=null, refAbstract=null), Reference(id=1329455753208901882, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Ma Shuai, Liu Jianwei, Zuo Xin. Survey on graph neural network[J]. Journal of Computer Research and Development, 2022, 59(1): 47−80(in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329455753284399355, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=李涵,严明玉,吕征阳,等. 图神经网络加速结构综述[J]. 计算机研究与发展,2021,58(6):1204−1229, articleTitle=null, refAbstract=null), Reference(id=1329455753364091133, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Han, Yan Mingyu, Lü Zhengyang, et al. Survey on graph neural network acceleration architectures[J]. Journal of Computer Research and Development, 2021, 58(6): 1204−1229(in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329455754991481087, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Velickovic P, Cucurull G, Casanova A, et al. Graph attention networks[J]. arXiv preprint, arXiv: 1710. 10903, 2017, articleTitle=null, refAbstract=null), Reference(id=1329455755121504512, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Ye Zhonglin, Zhao Haixing, Zhu Yu, et al. HSNR: A network representation learning algorithm using hierarchical structure embedding[J]. Chinese Journal of Electronics, 2019, 29(6): 1141−1152, articleTitle=null, refAbstract=null), Reference(id=1329455755197001985, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Kipf T N, Welling M. Variational graph auto-encoders[J]. arXiv preprint, arXiv: 1611. 07308, 2016, articleTitle=null, refAbstract=null), Reference(id=1329455755268305154, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Muhan, Chen Yixin. Link prediction based on graph neural networks [C] //Proc of the 32nd Int Conf on Neural Information Processing Systems. Red Hook: Curran Associates Inc, 2018: 5171–5181, articleTitle=null, refAbstract=null), Reference(id=1329455755331219715, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Ye Zhonglin, Zhao Haixing, Zhang Kecheng, et al. Tri-party deep network representation learning using inductive matrix completion[J]. Journal of Central South University, 2019, 26(10): 2746−2758, articleTitle=null, refAbstract=null), Reference(id=1329455755389939972, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Hamaguchi T, Oiwa H, Shimbo M, et al. Knowledge transfer for out-of-knowledge-base entities: A graph neural network approach[J]. arXiv preprint, arXiv: 1706. 05674, 2017, articleTitle=null, refAbstract=null), Reference(id=1329455755482214661, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Marcheggiani D, Bastings J, Titov I. Exploiting semantics in neural machine translation with graph convolutional networks[C] //Proc of the 16th Annual Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg, PA: ACL, 2018: 486−492, articleTitle=null, refAbstract=null), Reference(id=1329455755549323526, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Zonghan, Pan Shirui, Chen Fengwen, et al. A comprehensive survey on graph neural networks[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 32(1): 4−24, articleTitle=null, refAbstract=null), Reference(id=1329455755620626695, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Han Yi, Zhou Bin, Pei Jian, et al. Understanding importance of collaborations in co-authorship networks: A supportiveness analysis approach [C] //Proc of the 9th SIAM Int Conf on Data Mining. Philadelphia, PA: SIAM, 2009: 1112−1123, articleTitle=null, refAbstract=null), Reference(id=1329455755683541256, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Feng Yifan, You Haoxuan, Zhang Zizhao, et al. Hypergraph neural networks [C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 3558−3565, articleTitle=null, refAbstract=null), Reference(id=1329455755754844425, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Yadati N, Nimishakavi M, Yadav P, et al. HyperGCN: A new method of training graph convolutional networks on hypergraphs [C] // Proc of the 33rd Neural Information Processing Systems. New York: Curran Associates Inc, 2019: 1511−1522, articleTitle=null, refAbstract=null), Reference(id=1329455755868090634, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Chaofan, Cheng Zelei, Li Zuotian, et al. Hypergraph attention networks[C] //Proc of the 19th Int Conf on Trust, Security and Privacy in Computing and Communications (TrustCom). Piscataway, NJ: IEEE, 2020: 1560−1565, articleTitle=null, refAbstract=null), Reference(id=1329455755943588107, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Nong Liping, Wang Junyi, Lin Jiming, et al. Hypergraph wavelet neural networks for 3D object classification[J]. Neurocomputing, 2021, 463: 580−595, articleTitle=null, refAbstract=null), Reference(id=1329455756048445708, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Jianling, Ding Kaize, Hong Liangjie, et al. Next-item recommendation with sequential hypergraphs [C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 1101−1110, articleTitle=null, refAbstract=null), Reference(id=1329455756123943181, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Xu, Xiong Kun, Zhang Yongfeng, et al. Neural feature-aware recommendation with signed hypergraph convolutional network[J]. ACM Transactions on Information Systems, 2020, 39(1): 1−22, articleTitle=null, refAbstract=null), Reference(id=1329455756241383694, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Ruochi, Zou Yuesong, Ma Jian. Hyper-SAGNN: A self-attention based graph neural network for hypergraphs[J]. arXiv preprint, arXiv: 1911. 02613, 2019, articleTitle=null, refAbstract=null), Reference(id=1329455756321075471, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Yadati N, Nitin V, Nimishakavi M, et al. NHP: Neural hypergraph link prediction [C] //Proc of the 29th ACM Int Conf on Information & Knowledge Management. New York: ACM, 2020: 1705−1714, articleTitle=null, refAbstract=null), Reference(id=1329455756409155856, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Jie, Cui Ganqu, Hu Shengding, et al. Graph neural networks: A review of methods and applications[J]. AI Open, 2020, 1: 57−81, articleTitle=null, refAbstract=null), Reference(id=1329455756493041937, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=周飞燕,金林鹏,董军. 卷积神经网络研究综述[J]. 计算机学报,2017,40(6):1229−1251, articleTitle=null, refAbstract=null), Reference(id=1329455756560150802, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Feiyan, Jin Linpeng, Dong Jun. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017, 40(6): 1229−1251(in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329455756644036883, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Asif N A, Sarker Y, Chakrabortty P K, et al. Graph neural network: A comprehensive review on non-Euclidean space[J]. IEEE Access, 2021, 9: 60588−60606, articleTitle=null, refAbstract=null), Reference(id=1329455756715340052, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Malekzadeh M, Hajibabaee P, Heidari M, et al. Review of graph neural network in text classification [C] //Proc of the 12th Annual Ubiquitous Computing, Electronics & Mobile Communication Conf. Piscataway, NJ: IEEE, 2021: 0084−0091, articleTitle=null, refAbstract=null), Reference(id=1329455756786643221, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Shiwen, Zhang Wentao, Sun Fei, et al. Graph neural networks in recommender systems: A survey[J]. arXiv preprint, arXiv: 2011. 02260, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455756849557782, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Yue, Zhang Zizhao, Lin Haojie, et al. Hypergraph learning: Methods and practices[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 44(5): 2548−2566, articleTitle=null, refAbstract=null), Reference(id=1329455756916666647, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Yikang, Ouyang Wanli, Zhou Bolei, et al. Factorizable net: An efficient subgraph-based framework for scene graph generation[J]. arXiv preprint, arXiv: 1806. 11538, 2018, articleTitle=null, refAbstract=null), Reference(id=1329455756975386904, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Yan Sijie, Xiong Yuanjun, Lin Dahua. Spatial temporal graph convolutional networks for skeleton-based action recognition[C] //Proc of the 32nd AAAI Conf on Artificial Intelligence. Palo Alto, CA : AAAI, 2018: 7444–7452, articleTitle=null, refAbstract=null), Reference(id=1329455757046690073, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Yue, Sun Yongbin, Liu Ziwei, et al. Dynamic graph cnn for learning on point clouds[J]. ACM Transactions on Graphics, 2019, 38(5): 1−12, articleTitle=null, refAbstract=null), Reference(id=1329455757105410330, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Bastings J, Titov I, Aziz W, et al. Graph convolutional encoders for syntax-aware neural machine translation [C]// Proc of the 2017 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA : ACL, 2017: 1957−1967, articleTitle=null, refAbstract=null), Reference(id=1329455757168324891, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Beck D, Haffari G, Cohn T. Graph-to-sequence learning using gated graph neural networks[C] //Proc of the 56th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2018: 273−283, articleTitle=null, refAbstract=null), Reference(id=1329455757235433756, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu Ting, Yin Haoteng, Zhu Zhanxing. Spatiotemporal graph convolutional networks: A deep learning framework for traffic forecasting [C] //Proc of the 27th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2018: 3634−3640, articleTitle=null, refAbstract=null), Reference(id=1329455757306736925, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Guo Shengnan, Lin Youfang, Feng Ning. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting [C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA : AAAI , 2019: 922−929, articleTitle=null, refAbstract=null), Reference(id=1329455757382234398, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Zheng Chuanpan, Fan Xiaoliang, Wang Cheng, et al. GMAN: A graph multi-attention network for traffic prediction [C] //Proc of the 34th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 1234−1241, articleTitle=null, refAbstract=null), Reference(id=1329455757453537567, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Ying Rex, He Ruining, Chen Kaifeng, et al. Graph convolutional neural networks for web-scale recommender systems[C] //Proc of the 24th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2018: 974−983, articleTitle=null, refAbstract=null), Reference(id=1329455757512257824, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Fan Wenqi, Ma Yao, Li Qing, et al. Graph neural networks for social recommendation[C] //Proc of the 19th World Wide Web Conf. New York: ACM, 2019: 417−426, articleTitle=null, refAbstract=null), Reference(id=1329455757570978081, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Qitian, Zhang Hengrui, Gao Xiaofeng, et al. Dual graph attention networks for deep latent representation of multifaceted social effects in recommender systems[C] //Proc of the 19th World Wide Web. New York: ACM, 2019: 2091−2102, articleTitle=null, refAbstract=null), Reference(id=1329455757633892642, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=Zitnik M, Agrawal M, Leskovec J. Modeling polypharmacy side effects with graph convolutional networks[J]. Bioinformatics, 2018, 34(13): i457−i466, articleTitle=null, refAbstract=null), Reference(id=1329455757730361635, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Nuo, Wang Pinghui, Chen Long, et al. MR-GNN: Multi- resolution and dual graph neural network for predicting structured entity interactions[C] //Proc of the 28th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2019: 3968−3974, articleTitle=null, refAbstract=null), Reference(id=1329455757839413540, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=49, authorNames=null, journalName=null, refType=null, unstructuredReference=Do K, Tran T, Venkatesh S. Graph transformation policy network for chemical reaction prediction [C] //Proc of the 25th ACM SIGKDD Inter Conf on Knowledge Discovery & Data Mining. New York: ACM, 2019: 750−760, articleTitle=null, refAbstract=null), Reference(id=1329455759450026278, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=50, authorNames=null, journalName=null, refType=null, unstructuredReference=Schlichtkrull M, Kipf T, Bloem P, et al. Modeling relational data with graph convolutional networks [C] //Proc of the European Semantic Web Conf 2018. Berlin: Springer, 2018: 593−607, articleTitle=null, refAbstract=null), Reference(id=1329455759592632615, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=51, authorNames=null, journalName=null, refType=null, unstructuredReference=Chao Shang, Tang Yun, Huang Jing, et al. End-to-end structure-aware convolutional networks for knowledge base completion [C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 3060−3067, articleTitle=null, refAbstract=null), Reference(id=1329455759659741480, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=49, rfOrder=52, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Fanjin, Liu Xinyu, Tang Jie, et al, OAG: Toward linking large-scale heterogeneous entity graphs [C]//Proc of the 25th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2019: 2585−2595, articleTitle=null, refAbstract=null), Reference(id=1329455759722656041, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=50, rfOrder=53, authorNames=null, journalName=null, refType=null, unstructuredReference=Bretto A. Hypergraph Theory [M]. Berlin: Springer, 2013: 1−42, articleTitle=null, refAbstract=null), Reference(id=1329455759798153514, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=51, rfOrder=54, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Junjie, Zhao Xibin, Hu Han, et al. Emotion recognition from physiological signals using multi-hypergraph neural Networks[C] //Proc of the 2019 IEEE Int Conf on Multimedia and Expo. Piscataway, NJ: IEEE, 2019: 610−615, articleTitle=null, refAbstract=null), Reference(id=1329455759877845291, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=52, rfOrder=55, authorNames=null, journalName=null, refType=null, unstructuredReference=Jiang Jianwen, Wei Yuxuan, Feng Yifan, et al. Dynamic hypergraph neural networks [C] //Proc of the 28th Int Joint Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 2635−2641, articleTitle=null, refAbstract=null), Reference(id=1329455759944954156, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=53, rfOrder=56, authorNames=null, journalName=null, refType=null, unstructuredReference=Shao Jingzhi, Zhu Junjie, Wei Yuxuan, et al. Emotion recognition by edge-weighted hypergraph neural network [C] //Proc of the 2019 IEEE Int Conf on Image Processing. Piscataway, NJ: IEEE, 2019: 2144−2148, articleTitle=null, refAbstract=null), Reference(id=1329455760012063021, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=54, rfOrder=57, authorNames=null, journalName=null, refType=null, unstructuredReference=Fu Sichao, Liu Weifeng, Zhou Yicong et al. HpLapGCN: Hypergraph p-Laplacian graph convolutional networks[J]. Neurocomputing, 2019, 362: 166−174, articleTitle=null, refAbstract=null), Reference(id=1329455760074977582, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=55, rfOrder=58, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Kesu, Chen Jing, Liao Shijie, et al. Geographic-semantic- temporal hypergraph convolutional network for traffic flow prediction [C] //Proc of the 25th Intl Conf on Pattern Recognition. Piscataway, NJ: IEEE, 2021: 5444−5450, articleTitle=null, refAbstract=null), Reference(id=1329455760142086447, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=56, rfOrder=59, authorNames=null, journalName=null, refType=null, unstructuredReference=Bandyopadhyay S, Das K, Murty M N, Line hypergraph convolution network: Applying graph convolution for hypergraphs [J]. arXiv preprint, arXiv: 2002.03392, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455760200806704, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=57, rfOrder=60, authorNames=null, journalName=null, refType=null, unstructuredReference=Tran L H, Tran L H. Directed hypergraph neural network[J]. arXiv preprint, arXiv: 2008. 03626, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455760284692785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=58, rfOrder=61, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Yubo, Wang Nan, Chen Yufeng, et al. Hypergraph label propagation network[C] // Proc of the 34th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 6885−6892, articleTitle=null, refAbstract=null), Reference(id=1329455760372773170, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=59, rfOrder=62, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Chaoqi, Wang Ruijie, Yao Shuochao, et al. Hypergraph learning with line expansion[J]. arXiv preprint, arXiv: 2005. 04843, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455760448270643, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=60, rfOrder=63, authorNames=null, journalName=null, refType=null, unstructuredReference=Kim E S, Kang W Y, On K W, et al. Hypergraph attention networks for multimodal learning [C] // Proc of 2020 IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2020: 14569−14578, articleTitle=null, refAbstract=null), Reference(id=1329455760532156724, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=61, rfOrder=64, authorNames=null, journalName=null, refType=null, unstructuredReference=Dong Yihe, Sawin W, Bengio Y. HNHN: Hypergraph networks with hyperedge neurons[J]. arXiv preprint, arXiv: 2006. 12278, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455760599265589, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=62, rfOrder=65, authorNames=null, journalName=null, refType=null, unstructuredReference=Han Jiale, Cheng Bo, Wang Xu. Two-phase hypergraph based reasoning with dynamic relations for multi-hop KBQA [C] //Proc of the 29th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2021: 3615–3621, articleTitle=null, refAbstract=null), Reference(id=1329455760695734582, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=63, rfOrder=66, authorNames=null, journalName=null, refType=null, unstructuredReference=Te Gusi, Hu Wei, Guo Zongming, et al. Exploring hypergraph representation on face anti-spoofing beyond 2D attacks[C] //Proc of the 2020 IEEE Int Conf on Multimedia and Expo. Piscataway, NJ: IEEE , 2020[2021-02-01].https://ieeexplore.ieee.org/document/9102720/, articleTitle=null, refAbstract=null), Reference(id=1329455760775426359, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=64, rfOrder=67, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Shengyuan, Lv Pei, Zhang Yuzhen, et al. Semi-dynamic hypergraph neural network for 3D pose estimation [C] //Proc of the 29th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2020: 782−788, articleTitle=null, refAbstract=null), Reference(id=1329455760850923832, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=65, rfOrder=68, authorNames=null, journalName=null, refType=null, unstructuredReference=Fatemi B, Taslakian P, Vázquez D, et al. Knowledge hypergraphs: Prediction beyond binary relations [C] //Proc of the 29th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2020: 2191−2197, articleTitle=null, refAbstract=null), Reference(id=1329455760922227001, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=66, rfOrder=69, authorNames=null, journalName=null, refType=null, unstructuredReference=Yi J, Park J. Hypergraph convolutional recurrent neural network [C] //Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2020: 3366−3376, articleTitle=null, refAbstract=null), Reference(id=1329455761043861818, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=67, rfOrder=70, authorNames=null, journalName=null, refType=null, unstructuredReference=Ji Shuyi, Feng Yifan, Ji Rongrong, et al. Dual channel hypergraph collaborative filtering [C] //Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2020: 2020−2029, articleTitle=null, refAbstract=null), Reference(id=1329455761106776379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=68, rfOrder=71, authorNames=null, journalName=null, refType=null, unstructuredReference=Lostar M, Rekik I. Deep hypergraph U-Net for brain graph embedding and classification[J]. arXiv preprint, arXiv: 2008. 13118, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455761173885244, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=69, rfOrder=72, authorNames=null, journalName=null, refType=null, unstructuredReference=Banka A, Buzi I, Rekik I. Multi-view brain hyperconnectome autoencoder for brain state classification [C] // Proc of the 3rd Int Workshop of Predictive Intelligence in Medicine. Berlin: Springer, 2020: 101−110, articleTitle=null, refAbstract=null), Reference(id=1329455761236799805, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=70, rfOrder=73, authorNames=null, journalName=null, refType=null, unstructuredReference=Yadati N. Neural message passing for multi-relational ordered and recursive hypergraphs [C]//Proc of the 34th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2020: 3275–3289, articleTitle=null, refAbstract=null), Reference(id=1329455761299714366, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=71, rfOrder=74, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Xiangping, Chen Qingcai, Li Wei, et al. AdaHGNN: Adaptive hypergraph neural networks for multi-Label image classification [C] //Proc of the 28th ACM Inter Conf on Multimedia. New York: ACM, 2020: 284−293, articleTitle=null, refAbstract=null), Reference(id=1329455761366823231, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=72, rfOrder=75, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Xiangguo, Yin Hongzhi, Liu Bo, et al. Heterogeneous hypergraph embedding for graph classification [C] //Proc of the 14th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2021: 725−733, articleTitle=null, refAbstract=null), Reference(id=1329455761425543488, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=73, rfOrder=76, authorNames=null, journalName=null, refType=null, unstructuredReference=Shao Shuai, Xu Rui, Wang Yanjiang, et al. SAHDL: Sparse attention hypergraph regularized dictionary learning[J]. arXiv preprint, arXiv: 2010.12416, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455761488458049, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=74, rfOrder=77, authorNames=null, journalName=null, refType=null, unstructuredReference=Arya D, Gupta D K, Rudinac S, et al. HyperSAGE: Generalizing inductive representation learning on hypergraphs[J]. arXiv preprint, arXiv: 2010. 04558, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455761555566914, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=75, rfOrder=78, authorNames=null, journalName=null, refType=null, unstructuredReference=Ding Kaize, Wang Jianling, Li Jundong, et al. Be more with less: Hypergraph attention networks for inductive text classification [C] //Proc of the 2020 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2020: 4927−4936, articleTitle=null, refAbstract=null), Reference(id=1329455761626870083, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=76, rfOrder=79, authorNames=null, journalName=null, refType=null, unstructuredReference=Sawhney R, Agarwal S, Wadhwa A, et al. Spatiotemporal hypergraph convolution network for stock movement forecasting [C] //Proc of the 2020 IEEE Int Conf on Data Mining. Piscataway, NJ: IEEE, 2020: 482−491, articleTitle=null, refAbstract=null), Reference(id=1329455761714950468, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=77, rfOrder=80, authorNames=null, journalName=null, refType=null, unstructuredReference=Tan Yaqi, Ma Zhongchen, Zhan Yongzhao, et al. Hypergraph induced graph convolutional network for multi-label image recognition[C/OL] //Proc of the 2020 Int Conf on Internet of Things and Intelligent Applications. Piscataway, NJ: IEEE, 2020[2021-03-01].https://ieeexplore.ieee.org/document/9312371, articleTitle=null, refAbstract=null), Reference(id=1329455761786253637, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=78, rfOrder=81, authorNames=null, journalName=null, refType=null, unstructuredReference=Hou Rui, Small M, Forrest A R R. Community detection in a weighted directed hypergraph representation of cell-to-cell communication networks [J/OL]. bioRxiv, 2020[2021-03-01].https://doi.org/10.1101/ 2020.11.16.381566, articleTitle=null, refAbstract=null), Reference(id=1329455761861751110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=79, rfOrder=82, authorNames=null, journalName=null, refType=null, unstructuredReference=Bandyopadhyay S, Das K, Murty M N, Hypergraph attention isomorphism network by learning line graph expansion [C] //Proc of the 2020 IEEE Int Conf on Big Data. Piscataway, NJ: IEEE, 2020: 669−678, articleTitle=null, refAbstract=null), Reference(id=1329455761933054279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=80, rfOrder=83, authorNames=null, journalName=null, refType=null, unstructuredReference=Garasuie M M, Shabankhah M, Kamandi A. Improving hypergraph attention and hypergraph convolution networks [C] //Proc of the 11th Int Conf on Information and Knowledge Technology. Piscataway, NJ: IEEE, 2020: 67−72, articleTitle=null, refAbstract=null), Reference(id=1329455761995968840, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=81, rfOrder=84, authorNames=null, journalName=null, refType=null, unstructuredReference=Xia Xin, Yin Hongzhi, Yu Junliang, et al. Self-supervised hypergraph convolutional networks for session-based recommendation [C]// Proc of the 35th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2021: 4503−4511, articleTitle=null, refAbstract=null), Reference(id=1329455762109215049, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=82, rfOrder=85, authorNames=null, journalName=null, refType=null, unstructuredReference=Sawhney R, Agarwal S, Wadhwa A, et al. Stock selection via spatiotemporal hypergraph attention network: A learning to rank approach [C] //Proc of the 35th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2021: 497−504, articleTitle=null, refAbstract=null), Reference(id=1329455762197295434, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=83, rfOrder=86, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu Junliang, Yin Hongzhi, Li Jundong, et al. Self-supervised multi-channel hypergraph convolutional network for social recommendation [C] //Proc of the 21st World Wide Web. New York: ACM, 2021: 413−424, articleTitle=null, refAbstract=null), Reference(id=1329455762264404299, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=84, rfOrder=87, authorNames=null, journalName=null, refType=null, unstructuredReference=Hao Xiaoke, Li Jie, Yingchun, et al. Hypergraph neural network for skeleton-based action recognition[J]. IEEE Transactions on Image Processing, 2021, 30: 2263−2275, articleTitle=null, refAbstract=null), Reference(id=1329455762327318860, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=85, rfOrder=88, authorNames=null, journalName=null, refType=null, unstructuredReference=Xue Hansheng, Yang Luwei, Rajan V, et al. Multiplex bipartite network embedding using dual hypergraph convolutional networks [C] //Proc of the 21st World Wide Web. New York: ACM, 2021: 1649−1660, articleTitle=null, refAbstract=null), Reference(id=1329455763933737293, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=86, rfOrder=89, authorNames=null, journalName=null, refType=null, unstructuredReference=Maleki S, Wall D P, Pingali K. NetVec: A dcalable hypergraph embedding system[J]. arXiv preprint, arXiv: 2103. 09660v1, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455764009234766, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=87, rfOrder=90, authorNames=null, journalName=null, refType=null, unstructuredReference=Tudisco F, Prokopchik K, Benson A R. A nonlinear diffusion method for semi-supervised learning on hypergraphs[J]. arXiv preprint, arXiv: 2103. 14867, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455764097315151, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=88, rfOrder=91, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Jingcheng, Zhang Yong, Wei Yun, et al. Metro passenger flow prediction via dynamic hypergraph convolution networks[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(12): 7891−7903, articleTitle=null, refAbstract=null), Reference(id=1329455764177006928, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=89, rfOrder=92, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Binghao, Zhao Pengpeng, Zhuang Fuzhen, et al. Knowledge-aware hypergraph neural network for recommender systems[C] //Proc of the 26th Int Conf on Database Systems for Advanced Applications. Berlin: Springer, 2021: 132−147, articleTitle=null, refAbstract=null), Reference(id=1329455764302836049, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=90, rfOrder=93, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Jianling, Ding Kaize, Zhu Ziwei, et al. Session-based recommendation with hypergraph attention networks[J]. arXiv preprint, arXiv: 2112. 14266, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455764390916434, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=91, rfOrder=94, authorNames=null, journalName=null, refType=null, unstructuredReference=Bai Junjie, Gong Biao, Zhao Yining, et al. Multi-scale representation learning on hypergraph for 3D shape retrieval and recognition[J]. IEEE Transactions on Image Processing, 2021, 30: 5327−5338, articleTitle=null, refAbstract=null), Reference(id=1329455764462219603, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=92, rfOrder=95, authorNames=null, journalName=null, refType=null, unstructuredReference=Du Boxin, Yuan Changhe, Robert B, et al. Hypergraph pre-training with graph neural networks[J]. arXiv preprint, arXiv: 2105. 10862, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455764550299988, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=93, rfOrder=96, authorNames=null, journalName=null, refType=null, unstructuredReference=Huang Jing, Huang Xiaolin, Yang Jie. Residual enhanced multi- hypergraph neural network [C] // Proc of the 2021 IEEE Int Conf on Image Processing. Piscataway, NJ: IEEE, 2021: 3657−3661, articleTitle=null, refAbstract=null), Reference(id=1329455764617408853, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=94, rfOrder=97, authorNames=null, journalName=null, refType=null, unstructuredReference=Fu Jun, Hou Chengbin, Zhou Wei, et al. Adaptive hypergraph convolutional network for no-reference 360-degree image quality assessment [C] //Proc of the 30th ACM Inte Conf on Multimedia. New York: ACM, 2021: 961–969, articleTitle=null, refAbstract=null), Reference(id=1329455764688712022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=95, rfOrder=98, authorNames=null, journalName=null, refType=null, unstructuredReference=Huang Jing, Yang Jie. UniGNN: A unified framework for graph and hypergraph neural networks [C] //Proc of the 30th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2021: 2563−2569, articleTitle=null, refAbstract=null), Reference(id=1329455764760015191, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=96, rfOrder=99, authorNames=null, journalName=null, refType=null, unstructuredReference=Jo J, Baek Ji, Lee S, et al. Edge representation learning with hypergraphs[J]. arXiv preprint, arXiv: 2106. 15845, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455764818735448, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=97, rfOrder=100, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Jiying, Chen Yuzhao, Xiao Xiong, et al. Learnable hypergraph Laplacian for hypergraph learning [C] //Proc of the 2022 IEEE Int Conf on Acoustics, Speech and Signal Processing. Piscataway, NJ: IEEE, 2022: 4503−4507, articleTitle=null, refAbstract=null), Reference(id=1329455764885844313, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=98, rfOrder=101, authorNames=null, journalName=null, refType=null, unstructuredReference=Yan Yichao, Qin Jie, Chen Jiaxin, et al. Learning multi-granular hypergraphs for video-based person re-identification [C] //Proc of the 2020 IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2020: 2896−2905, articleTitle=null, refAbstract=null), Reference(id=1329455764944564570, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=99, rfOrder=102, authorNames=null, journalName=null, refType=null, unstructuredReference=Chien E, Pan Chao, Peng Jianhao, et al. You are AllSet: A multiset function framework for hypergraph neural networks[J]. arXiv preprint, arXiv: 2106. 13264, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455765011673435, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=100, rfOrder=103, authorNames=null, journalName=null, refType=null, unstructuredReference=Cui Chaoran, Li Xiaojie, Du Juan, et al. Temporal-relational hypergraph tri-attention networks for stock trend prediction[J]. arXiv preprint, arXiv: 2107. 14033, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455765091365212, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=101, rfOrder=104, authorNames=null, journalName=null, refType=null, unstructuredReference=Pan Junren, Lei Baiying, Shen Yanyan, et al. Characterization multimodal connectivity of brain network by hypergraph GAN for Alzheimer’s disease analysis [C] //Proc of the 4th Chinese Conf on Pattern Recognition and Computer Vision. Berlin: Springer, 2021: 467−478, articleTitle=null, refAbstract=null), Reference(id=1329455765158474077, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=102, rfOrder=105, authorNames=null, journalName=null, refType=null, unstructuredReference=Zuo Qinkun, Lei Baiying, Shen Yanyan, et al. Multimodal representations learning and adversarial hypergraph fusion for early Alzheimer’s disease prediction [C] //Proc of the 4th Chinese Conf on Pattern Recognition and Computer Vision. Berlin: Springer, 2021: 479− 490, articleTitle=null, refAbstract=null), Reference(id=1329455765225582942, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=103, rfOrder=106, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Yichao, Chen Hongxu, Sun Xiangguo, et al. Hyperbolic hypergraphs for sequential recommendation [C] //Proc of the 30th ACM Int Conf on Information & Knowledge Management. New York: ACM, 2021: 988−997, articleTitle=null, refAbstract=null), Reference(id=1329455765301080415, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=104, rfOrder=107, authorNames=null, journalName=null, refType=null, unstructuredReference=Yadati N, Gao T, Asoodeh S, et al. Graph neural networks for soft semi-supervised learning on hypergraphs [C] //Proc of the 25th Pacific-Asia Conf on Knowledge Discovery and Data Mining. Berlin: Springer, 2021: 447−458, articleTitle=null, refAbstract=null), Reference(id=1329455765372383584, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=105, rfOrder=108, authorNames=null, journalName=null, refType=null, unstructuredReference=Ma Zhongtian, Jiang Zhiguo, Zhang Haopeng. Hyperspectral image classification using spectral-spatial hypergraph convolution neural network[J]. SPIE Remote Sensing, 2021, 11862: 118620I, articleTitle=null, refAbstract=null), Reference(id=1329455765443686753, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=106, rfOrder=109, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Longcan, Wang Daling, Song Kaisong, et al. Dual-view hypergraph neural networks for attributed graph learning[J]. Knowledge-Based Systems, 2021, 227: 107185, articleTitle=null, refAbstract=null), Reference(id=1329455765506601314, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=107, rfOrder=110, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Junwei, Gao Min, Yu Junliang, et al. Double-scale self- supervised hypergraph learning for group recommendation[C] //Proc of the 30th ACM Int Conf on Information & Knowledge Management. New York: ACM, 2021: 2557−2567, articleTitle=null, refAbstract=null), Reference(id=1329455765586293091, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=108, rfOrder=111, authorNames=null, journalName=null, refType=null, unstructuredReference=Vijaikumar M, Hada D V, Shevade S K. HyperTeNet: Hypergraph and transformer-based neural network for personalized list continuation [C] //Proc of the 2021 IEEE Int Conf on Data Mining. Piscataway, NJ: IEEE 2021: 1210−1215, articleTitle=null, refAbstract=null), Reference(id=1329455765657596260, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=109, rfOrder=112, authorNames=null, journalName=null, refType=null, unstructuredReference=Xia Zhongxiu, Zhang Weiyu, Weng Ziqiang. Social recommendation system based on hypergraph attention network[J]. Computational Intelligence and Neuroscience, 2021, 2021: 7716214, articleTitle=null, refAbstract=null), Reference(id=1329455765724705125, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=110, rfOrder=113, authorNames=null, journalName=null, refType=null, unstructuredReference=Ding Meirong, Lin Xiaokang, Zeng Biqing, et al. Hypergraph neural networks with attention mechanism for session-based recommendation[J]. Journal of Physics: Conference Series, 2021, 2082(1): 012007, articleTitle=null, refAbstract=null), Reference(id=1329455765783425382, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=111, rfOrder=114, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Zirui, Gao Chen, Chen Xu, et al. Inhomogeneous social recommendation with hypergraph convolutional networks[J]. arXiv preprint, arXiv: 2111. 03344, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455765850534247, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=112, rfOrder=115, authorNames=null, journalName=null, refType=null, unstructuredReference=Luo Xiaoyi, Peng Jiaheng, Liang Jun. Directed hypergraph attention network for traffic forecasting[J]. IET Intelligent Transport Systems, 2022, 16(1): 85−98, articleTitle=null, refAbstract=null), Reference(id=1329455765913448808, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=113, rfOrder=116, authorNames=null, journalName=null, refType=null, unstructuredReference=Lin Jingjing, Ye Zhonglin, Zhao Haixing, et al. DeepHGNN: A novel deep hypergraph neural network[J]. Chinese Journal of Electronics, 2022, 31(5): 958−968, articleTitle=null, refAbstract=null), Reference(id=1329455765972169065, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=114, rfOrder=117, authorNames=null, journalName=null, refType=null, unstructuredReference=Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks [J]. arXiv preprint, arXiv:1609. 02907, 2017, articleTitle=null, refAbstract=null), Reference(id=1329455766047666538, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=115, rfOrder=118, authorNames=null, journalName=null, refType=null, unstructuredReference=Bruna J, Zaremba W, Szlam A, et al. Spectral networks and locally connected networks on graphs[J]. arXiv preprint, arXiv: 1312. 6203, 2014, articleTitle=null, refAbstract=null), Reference(id=1329455766135746923, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=116, rfOrder=119, authorNames=null, journalName=null, refType=null, unstructuredReference=Defferrard M, Bresson X, Vandergheynst P. Convolutional neural networks on graphs with fast localized spectral filtering[C] //Proc of the 30th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2016: 3844−3852, articleTitle=null, refAbstract=null), Reference(id=1329455766202855788, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=117, rfOrder=120, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Dengyong, Huang Jiayuan, Schölkopf B. Learning with hypergraphs: Clustering, classification, and embedding [C] // Proc of the 19th Int Conf on Neural Information Processing Systems. Cambridge, MA: MIT, 2006: 1601–1608, articleTitle=null, refAbstract=null), Reference(id=1329455766286741869, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=118, rfOrder=121, authorNames=null, journalName=null, refType=null, unstructuredReference=Song Bai, Zhang Feihu, Torr P H S. Hypergraph convolution and hypergraph attention[J]. Pattern Recognition, 2021, 110: 107637, articleTitle=null, refAbstract=null), Reference(id=1329455766358045038, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=119, rfOrder=122, authorNames=null, journalName=null, refType=null, unstructuredReference=Ma Xueqi, Liu Weifeng, Li Shuying, et al. Hypergraph p-Laplacian regularization for remotely sensed image recognition[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(3): 1585−1595, articleTitle=null, refAbstract=null), Reference(id=1329455766420959599, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=120, rfOrder=123, authorNames=null, journalName=null, refType=null, unstructuredReference=Donnat C, Zitnik M, Hallac D, et al. Learning structural node embeddings via diffusion wavelets[J]. arXiv preprint, arXiv: 1710. 10321, 2018, articleTitle=null, refAbstract=null), Reference(id=1329455766496457072, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=121, rfOrder=124, authorNames=null, journalName=null, refType=null, unstructuredReference=Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift [C] //Proc of the 32nd Inte Conf on Machine Learning. New York: ACM, 2015: 448−456, articleTitle=null, refAbstract=null), Reference(id=1329455766567760241, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=122, rfOrder=125, authorNames=null, journalName=null, refType=null, unstructuredReference=Louis A. Hypergraph Markov operators, eigenvalues and approximation algorithms [C] //Proc of the 47th Annual ACM Symp on Theory of Computing. New York: ACM, 2015: 713−722, articleTitle=null, refAbstract=null), Reference(id=1329455766639063410, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=123, rfOrder=126, authorNames=null, journalName=null, refType=null, unstructuredReference=Chan T H H, Louis A, Tang Z G, et al. Spectral properties of hypergraph Laplacian and approximation algorithms[J]. Journal of the ACM, 2018, 65(3): 1−48, articleTitle=null, refAbstract=null), Reference(id=1329455766706172275, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=124, rfOrder=127, authorNames=null, journalName=null, refType=null, unstructuredReference=Chan T H H, Liang Zhibin. Generalizing the hypergraph Laplacian via a diffusion process with mediators [C] // Proc of the 24th Int Computing and Combinatorics Computing and Combinatorics. Berlin: Springer, 2020: 441–453, articleTitle=null, refAbstract=null), Reference(id=1329455766764892532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=125, rfOrder=128, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Keyulu, Hu Weihua, Leskovec J, et al. How powerful are graph neural networks?[J]. arXiv preprint, arXiv: 1810. 00826, 2019, articleTitle=null, refAbstract=null), Reference(id=1329455768409059702, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=126, rfOrder=129, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Xiangguo, Yin Hongzhi, Liu Bo, et al. Multi-level hyperedge distillation for social linking prediction on sparsely observed networks[C] //Proc of the 21st World Wide Web Conf. New York: ACM, 2021: 2934−2945, articleTitle=null, refAbstract=null), Reference(id=1329455768484557176, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=127, rfOrder=130, authorNames=null, journalName=null, refType=null, unstructuredReference=Xia Liqiao, Zheng Pai, Huang Xiao, et al. A novel hypergraph convolution network-based approach for predicting the material removal rate in chemical mechanical planarization[J]. Journal of Intelligent Manufacturing, 2021, 33(8): 2295−2306, articleTitle=null, refAbstract=null), Reference(id=1329455768555860345, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=128, rfOrder=131, authorNames=null, journalName=null, refType=null, unstructuredReference=Alfke D, Stoll M. Semi-supervised classification on non-sparse graphs using low-rank graph convolutional networks[J]. arXiv preprint, arXiv: 1905. 10224v1, 2019, articleTitle=null, refAbstract=null), Reference(id=1329455768643940730, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=129, rfOrder=132, authorNames=null, journalName=null, refType=null, unstructuredReference=Chami I, Rex Y, Ré C, et al. Hyperbolic graph convolutional neural networks [J]. arXiv preprint, arXiv : 1910.12933, 2019, articleTitle=null, refAbstract=null), Reference(id=1329455768702660987, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=130, rfOrder=133, authorNames=null, journalName=null, refType=null, unstructuredReference=Srinivasan B, Zheng Da, Karypis G. Learning over families of sets-hypergraph representation learning for higher order tasks[J]. arXiv preprint, arXiv: 2101. 07773v1, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455768761381244, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=131, rfOrder=134, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Ming, Wei Zhewei, Huang Zengfeng, et al. Simple and deep graph convolutional networks[C] //Proc of the 37th Int Conf on Machine Learning. New York: ACM, 2020: 1725−1735, articleTitle=null, refAbstract=null), Reference(id=1329455768836878717, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=132, rfOrder=135, authorNames=null, journalName=null, refType=null, unstructuredReference=Payne J. Deep hyperedges: A framework for transductive and inductive learning on hypergraphs[J]. arXiv preprint, arXiv: 1910. 02633, 2019, articleTitle=null, refAbstract=null), Reference(id=1329455768979485054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=133, rfOrder=136, authorNames=null, journalName=null, refType=null, unstructuredReference=Gilmer J, Schoenholz S S, Riley P F, et al. Neural message passing for quantum chemistry [C] //Proc of the 34th Int Conf on Machine Learning. New York: ACM, 2017: 1263−1272, articleTitle=null, refAbstract=null), Reference(id=1329455769050788223, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=134, rfOrder=137, authorNames=null, journalName=null, refType=null, unstructuredReference=Hamilton W L, Ying R, Jure L. Inductive representation learning on large graphs[C] //Proc of the 31st Int Conf on Neural Information Processing Systems. Red Hook: Curran Associates Inc, 2017: 1025−1035, articleTitle=null, refAbstract=null), Reference(id=1329455769113702784, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=135, rfOrder=138, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Qiao, Zeng Yifu, Mokhosi R, et al. STAMP: Short-term attention/memory priority model for session-based recommendation [C] //Proc of the 24th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2018: 1831−1839, articleTitle=null, refAbstract=null), Reference(id=1329455769176617345, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=136, rfOrder=139, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Shu, Tang Yuyuan, Zhu Yanqiao, et al. Session-based recommendation with graph neural networks[C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 346–353, articleTitle=null, refAbstract=null), Reference(id=1329455769252114818, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=137, rfOrder=140, authorNames=null, journalName=null, refType=null, unstructuredReference=He Xiangnan, Deng Kuan, Wang Xiang, et al. LightGCN: Simplifying and powering graph convolution network for recommendation [C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 639−648, articleTitle=null, refAbstract=null), Reference(id=1329455769302446467, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=138, rfOrder=141, authorNames=null, journalName=null, refType=null, unstructuredReference=Wadhwa G, Dhall A, Murala S, et al. Hyperrealistic image inpainting with hypergraphs[C] //Proc of the 2021 IEEE Winter Conf on Applications of Computer Vision. Piscataway, NJ: IEEE, 2021: 3911−3920, articleTitle=null, refAbstract=null), Reference(id=1329455769365361028, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=139, rfOrder=142, authorNames=null, journalName=null, refType=null, unstructuredReference=Dang N T V, Tran L H, Tran L H. Noise-robust classification with hypergraph neural network[J]. arXiv preprint, arXiv: 2102. 01934, 2021, articleTitle=null, refAbstract=null), Reference(id=1329455769432469893, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=140, rfOrder=143, authorNames=null, journalName=null, refType=null, unstructuredReference=Zheng Wenbo, Yan Lan, Gou Chao, et al. Two heads are better than one: Hypergraph-enhanced graph reasoning for visual event ratiocination [C] //Proc of the 38th Int Conf on Machine Learning. New York: ACM, 139: 12747−12760, articleTitle=null, refAbstract=null), Reference(id=1329455769499578758, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=141, rfOrder=144, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhao Weizhong, Zhang Jinyong, Yang Jincai, et al. A novel joint biomedical event extraction framework via two-level modeling of documents[J]. Information Sciences, 2021, 550: 27−40, articleTitle=null, refAbstract=null), Reference(id=1329455769579270535, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=142, rfOrder=145, authorNames=null, journalName=null, refType=null, unstructuredReference=Vashishth S, Sanyal S, Nitin V, et al. Composition-based multi-relational graph convolutional networks[J]. arXiv preprint, arXiv: 1911. 038082, 2020, articleTitle=null, refAbstract=null), Reference(id=1329455769637990792, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=143, rfOrder=146, authorNames=null, journalName=null, refType=null, unstructuredReference=Yadati N, Dayanidhi R, Vaishnavi S, et al. Knowledge base question answering through recursive hypergraphs[C] //Proc of the 16th Conf of the European Chapter of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2021: 448−454, articleTitle=null, refAbstract=null), Reference(id=1329455769709293961, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=144, rfOrder=147, authorNames=null, journalName=null, refType=null, unstructuredReference=Han Jiale, Cheng Bo, Wang Xu. Open domain question answering based on text enhanced knowledge graph with hyperedge infusion [C] //Proc of the 2020 Findings of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2020: 1475−1481, articleTitle=null, refAbstract=null), Reference(id=1329455769776402826, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=145, rfOrder=148, authorNames=null, journalName=null, refType=null, unstructuredReference=Miller A, Fisch A, Dodge J, et al. Key-value memory networks for directly reading documents [C] //Proc of the 2016 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2016: 1400–1409, articleTitle=null, refAbstract=null), Reference(id=1329455769843511691, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=146, rfOrder=149, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Haitian, Dhingra B, Zaheer M, et al. Open domain question answering using early fusion of knowledge bases and text[J]. arXiv preprint, arXiv: 1809, 00782, 2018, articleTitle=null, refAbstract=null), Reference(id=1329455769910620556, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=147, rfOrder=150, authorNames=null, journalName=null, refType=null, unstructuredReference=Dongen S V. Graph clustering by flow simulation [D]. Amsterdam: Center for Math and Computer Science, 2000, articleTitle=null, refAbstract=null), Reference(id=1329455769977729421, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=148, rfOrder=151, authorNames=null, journalName=null, refType=null, unstructuredReference=Blondel V D, Guillaume J L, Lambiotte R, et al. Fast unfolding of communities in large networks[J]. Journal of Statistical Mechanics: Theory and Experiment, 2008, 2008(10): 10008, articleTitle=null, refAbstract=null), Reference(id=1329455770036449678, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=149, rfOrder=152, authorNames=null, journalName=null, refType=null, unstructuredReference=Madine M M, Rekik I, Werghi N. Diagnosing autism using T1-W MRI with multi-kernel learning and hypergraph neural network [C] //Proc of the 2020 IEEE Int Conf on Image Processing. Piscataway, NJ: IEEE, 2020: 438−442, articleTitle=null, refAbstract=null), Reference(id=1329455770145501583, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=150, rfOrder=153, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Keyulu, Li Chengtao, Tian Yonglong, et al. Representation learning on graphs with jumping knowledge networks[C] //Proc of the 35th Int Conf on Machine Learning. New York: ACM, 2018: 5453−5462, articleTitle=null, refAbstract=null), Reference(id=1329455770212610448, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=151, rfOrder=154, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu Rong, Huang Wenbing, Xu Tingyang, et al. DropEdge: Towards deep graph convolutional networks on node classification[J]. arXiv preprint, arXiv: 1907. 10903, 2020, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)]), AuthorCompany(id=1329455743322927261, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=3, ext=[AuthorCompanyExt(id=1329455743327121566, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743322927261, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Department of Information Engineering, Xining Urban Vocational & Technical College, Xining 810003), AuthorCompanyExt(id=1329455743335510175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743322927261, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3西宁城市职业技术学院信息工程系 西宁 810003)])], figs=[ArticleFig(id=1329455746967777476, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Figure 1, caption=The sequential development of hypergraph neural network model, figureFileSmall=6ujKIUyHn6gIfA0cqZo7iQ==, figureFileBig=zJaNaA/TKhRzd+xGCe4HfA==, tableContent=null), ArticleFig(id=1329455747043274949, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=图1, caption=超图神经网络模型的发展时序, figureFileSmall=6ujKIUyHn6gIfA0cqZo7iQ==, figureFileBig=zJaNaA/TKhRzd+xGCe4HfA==, tableContent=null), ArticleFig(id=1329455747273961670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Figure 2, caption=Hypergraph and graph, figureFileSmall=gUuRzCJGfUX59rU6eBBsHQ==, figureFileBig=SKRYrUmwYG6xqEnGty1opw==, tableContent=null), ArticleFig(id=1329455747391402183, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=图2, caption=超图和图, figureFileSmall=gUuRzCJGfUX59rU6eBBsHQ==, figureFileBig=SKRYrUmwYG6xqEnGty1opw==, tableContent=null), ArticleFig(id=1329455747567562952, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Figure 3, caption=Comparison of HGNN and MultiHGNN, figureFileSmall=XPzgdZ3Q1sExYPq/aQbZCQ==, figureFileBig=GwGbX6xecuPTbaWGO6pDuw==, tableContent=null), ArticleFig(id=1329455747710169289, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=图3, caption=HGNN与MultiHGNN的比较, figureFileSmall=XPzgdZ3Q1sExYPq/aQbZCQ==, figureFileBig=GwGbX6xecuPTbaWGO6pDuw==, tableContent=null), ArticleFig(id=1329455747848581322, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 1, caption=

The Symbol Description

, figureFileSmall=null, figureFileBig=null, tableContent=
符号含义
$ \mathcal{G} $超图
${\mathcal{G}_{\mathrm{D}}}$有向超图
$ {{{\boldsymbol{H}}_\mathcal{G}}} $超图的关联矩阵
$ {{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}} $有向超图的关联矩阵
$ \boldsymbol{H}_{\mathcal{G}_{\mathrm{D}}}^{\mathrm{ head}} $有向超图头关联矩阵
${{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{tail}}}}$有向超图尾关联矩阵
$ {{\boldsymbol{X}}_\mathcal{G}} $顶点的特征矩阵
${{\boldsymbol{x}}_i}$顶点$i$的特征
$L(\mathcal{G})$超图$\mathcal{G}$的线图
$ {{{\boldsymbol{A}}_{L(\mathcal{G})}}} $$\mathcal{G}$的线图$L(\mathcal{G})$的邻接矩阵
${\mathcal{L}_\mathcal{G}}$$\mathcal{G}$的拉普拉斯矩阵
${{\mathcal{L}}_{{\mathcal{G}_{\mathrm{D}}}}}$${\mathcal{G}_{\mathrm{D}}}$的拉普拉斯矩阵
${\boldsymbol{I}} $单位矩阵
$ \odot $哈达码乘积
${\boldsymbol{\sigma}} $, ${{\boldsymbol{\sigma}} _{{\text{att}}}}$激活函数
), ArticleFig(id=1329455747978604747, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表1, caption=

符号描述

, figureFileSmall=null, figureFileBig=null, tableContent=
符号含义
$ \mathcal{G} $超图
${\mathcal{G}_{\mathrm{D}}}$有向超图
$ {{{\boldsymbol{H}}_\mathcal{G}}} $超图的关联矩阵
$ {{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}} $有向超图的关联矩阵
$ \boldsymbol{H}_{\mathcal{G}_{\mathrm{D}}}^{\mathrm{ head}} $有向超图头关联矩阵
${{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{tail}}}}$有向超图尾关联矩阵
$ {{\boldsymbol{X}}_\mathcal{G}} $顶点的特征矩阵
${{\boldsymbol{x}}_i}$顶点$i$的特征
$L(\mathcal{G})$超图$\mathcal{G}$的线图
$ {{{\boldsymbol{A}}_{L(\mathcal{G})}}} $$\mathcal{G}$的线图$L(\mathcal{G})$的邻接矩阵
${\mathcal{L}_\mathcal{G}}$$\mathcal{G}$的拉普拉斯矩阵
${{\mathcal{L}}_{{\mathcal{G}_{\mathrm{D}}}}}$${\mathcal{G}_{\mathrm{D}}}$的拉普拉斯矩阵
${\boldsymbol{I}} $单位矩阵
$ \odot $哈达码乘积
${\boldsymbol{\sigma}} $, ${{\boldsymbol{\sigma}} _{{\text{att}}}}$激活函数
), ArticleFig(id=1329455748049907916, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 2, caption=

Fourier Transform and Wavelet Transform

, figureFileSmall=null, figureFileBig=null, tableContent=
傅里叶变换傅里叶逆变换小波变换及其
近似多项式
小波逆变换及其
近似多项式
${\boldsymbol{U}}_\mathcal{G}^{\mathrm{T}}$${{\boldsymbol{U}}_\mathcal{G}}$$ {({\boldsymbol{\psi}} _s^\mathcal{G})^{ - 1}} \approx (\overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G})' = \sum\limits_{k = 0}^{K'} {{{\overline {{\theta}} '}_k}{{({{\overline {\boldsymbol{\varTheta}} }^\mathcal{G}})}^k}} $${\boldsymbol{\psi}} _s^\mathcal{G} \approx \overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G} = \sum\limits_{k = 0}^K {{{\overline \theta }_k}{{({{\overline {{\boldsymbol{{\varTheta}}}} }^\mathcal{G}})}^k}} $
), ArticleFig(id=1329455748125405389, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表2, caption=

傅里叶变换和小波变换

, figureFileSmall=null, figureFileBig=null, tableContent=
傅里叶变换傅里叶逆变换小波变换及其
近似多项式
小波逆变换及其
近似多项式
${\boldsymbol{U}}_\mathcal{G}^{\mathrm{T}}$${{\boldsymbol{U}}_\mathcal{G}}$$ {({\boldsymbol{\psi}} _s^\mathcal{G})^{ - 1}} \approx (\overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G})' = \sum\limits_{k = 0}^{K'} {{{\overline {{\theta}} '}_k}{{({{\overline {\boldsymbol{\varTheta}} }^\mathcal{G}})}^k}} $${\boldsymbol{\psi}} _s^\mathcal{G} \approx \overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G} = \sum\limits_{k = 0}^K {{{\overline \theta }_k}{{({{\overline {{\boldsymbol{{\varTheta}}}} }^\mathcal{G}})}^k}} $
), ArticleFig(id=1329455748200902862, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 3, caption=

Common Datasets of Citation Network

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集节点数目边数目超边数目
Cora27085429
Cora(co-citation)20781579
Cora(co-authorship)20781072
Pubmed1971744338
Pubmed(co-citation)197177963
Citeseer33124732
Citeseer(co-citation)33121079
DBLP (co-authorship)4341322535
), ArticleFig(id=1329455748280594639, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表3, caption=

引文网络常用数据集

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集节点数目边数目超边数目
Cora27085429
Cora(co-citation)20781579
Cora(co-authorship)20781072
Pubmed1971744338
Pubmed(co-citation)197177963
Citeseer33124732
Citeseer(co-citation)33121079
DBLP (co-authorship)4341322535
), ArticleFig(id=1329455748351897808, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 4, caption=

Hypergraph Neural Network Models, Datasets and Evaluations Applied for Citation Network Classification

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
HGNNCora, PubmedAccuracy
DHGNNaCoraAccuracy
HyperGCNCora, DBLP, Pubmed, CiteseerMean test error ± Standard deviation, Density
LHCNCora, Citeseer, PubmedAccuracy
HAINCora, Citeseer, Pubmed, DBLPAccuracy
HWNNCora, Pubmed, DBLPRecall, F1, Accuracy, Precision
HpLapGCNCora, CiteseerAccuracy
Directed hypergraph neural network[57]Cora, CiteseerAccuracy
DHNCora, DBLP, ACM, Amazon, arXivMean squared error
MPNN-R[108]Cora, DBLP, ACM, arXivMean test error ± Standard deviation
HyperSAGECora, DBLP, Pubmed, CiteseerMean test error ± Standard deviation, Accuracy
HyperGeneCora, Pubmed, Corum, DisgenetAccuracy
HNHNCora, DBLP, Pubmed, CiteseerAccuracy
HERALDCoraAccuracy, Standard deviation
DHGNNbCora, Citeseer, DBLP, IMDBMicro-F1
iHGANCora, CiteseerAccuracy
HyperNDCora, Citeseer, DBLPAccuracy
), ArticleFig(id=1329455748427395281, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表4, caption=

应用于引文网络分类的超图神经网络模型、数据集和评价指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
HGNNCora, PubmedAccuracy
DHGNNaCoraAccuracy
HyperGCNCora, DBLP, Pubmed, CiteseerMean test error ± Standard deviation, Density
LHCNCora, Citeseer, PubmedAccuracy
HAINCora, Citeseer, Pubmed, DBLPAccuracy
HWNNCora, Pubmed, DBLPRecall, F1, Accuracy, Precision
HpLapGCNCora, CiteseerAccuracy
Directed hypergraph neural network[57]Cora, CiteseerAccuracy
DHNCora, DBLP, ACM, Amazon, arXivMean squared error
MPNN-R[108]Cora, DBLP, ACM, arXivMean test error ± Standard deviation
HyperSAGECora, DBLP, Pubmed, CiteseerMean test error ± Standard deviation, Accuracy
HyperGeneCora, Pubmed, Corum, DisgenetAccuracy
HNHNCora, DBLP, Pubmed, CiteseerAccuracy
HERALDCoraAccuracy, Standard deviation
DHGNNbCora, Citeseer, DBLP, IMDBMicro-F1
iHGANCora, CiteseerAccuracy
HyperNDCora, Citeseer, DBLPAccuracy
), ArticleFig(id=1329455748498698450, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 5, caption=

Experimental Results of Different Models on Cora Dataset %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型Accuracy
GCN81.5
HGNN81.6
DHGNNa82.5
iHGAN82.9
), ArticleFig(id=1329455748574195923, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表5, caption=

不同模型在Cora数据集上的实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型Accuracy
GCN81.5
HGNN81.6
DHGNNa82.5
iHGAN82.9
), ArticleFig(id=1329455748632916180, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 6, caption=

Common Datasets of Session-Based Recommendation

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集训练会话个数测试会话个数商品个数会话的
平均长度
YooChoose 1/4591774555898296185.71
Diginetica71947060858430975.12
), ArticleFig(id=1329455748691636437, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表6, caption=

基于会话推荐的常用数据集

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集训练会话个数测试会话个数商品个数会话的
平均长度
YooChoose 1/4591774555898296185.71
Diginetica71947060858430975.12
), ArticleFig(id=1329455748754550998, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 7, caption=

Results of Different Models for Session-Based Recommendation(MRR@K) %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集
YooChoose (1/4)Diginetica
STAMP30.0014.32
SR-GNN31.8917.59
HGNNA30.8117.74
SHARE32.1118.05
DHCN36.7230.58
S2-DHCN[62]40.1330.94
), ArticleFig(id=1329455748830048471, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表7, caption=

基于会话推荐不同模型的结果(MRR@K)

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集
YooChoose (1/4)Diginetica
STAMP30.0014.32
SR-GNN31.8917.59
HGNNA30.8117.74
SHARE32.1118.05
DHCN36.7230.58
S2-DHCN[62]40.1330.94
), ArticleFig(id=1329455748897157336, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 8, caption=

Common Datasets of Social Recommendation

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集用户个数商品个数交互个数关系个数
LastFM1892176329284325434
Douban28483958689488735770
Yelp1953921266450884363672
), ArticleFig(id=1329455748976849113, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表8, caption=

社会推荐的常用数据集

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集用户个数商品个数交互个数关系个数
LastFM1892176329284325434
Douban28483958689488735770
Yelp1953921266450884363672
), ArticleFig(id=1329455750604239067, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 9, caption=

Results of Different Models for Social Recommendation %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集Recall@KPrecision@KNDCG@K
DHCFLastFM17.13116.87720.744
LightGCN19.48019.20523.392
MHCN19.94519.62523.834
HASRE20.21919.86724.084
DHCFDouban5.75516.87118.655
LightGCN6.24717.78019.881
MHCN5.66518.28320.694
HASRE6.98218.81721.699
DHCFYelp5.9862.2984.700
LightGCN6.5252.5864.998
MHCN6.8622.7515.356
HASRE7.1222.8655.691
), ArticleFig(id=1329455750688125148, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表9, caption=

社会推荐不同模型的结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集Recall@KPrecision@KNDCG@K
DHCFLastFM17.13116.87720.744
LightGCN19.48019.20523.392
MHCN19.94519.62523.834
HASRE20.21919.86724.084
DHCFDouban5.75516.87118.655
LightGCN6.24717.78019.881
MHCN5.66518.28320.694
HASRE6.98218.81721.699
DHCFYelp5.9862.2984.700
LightGCN6.5252.5864.998
MHCN6.8622.7515.356
HASRE7.1222.8655.691
), ArticleFig(id=1329455750826537181, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 10, caption=

Hypergraph Neural Network Models, Datasets and Evaluation Applied for Recommendation System

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
HyperRecAmazon, Etsy, GoodreadsHits@K, NDCG@K, MRR
H2SeqRecAMT, GoodreadsHR@K, NDCG@K
DHCFMovieLens-100K, CiteUlike-A, Baidu-Feed, Baidu-newsNDCG@K, Precision@K, Recall@K, Hits@K
SHCNAmazon, YelpHR, F1, NDCG
DHCNYoochoose, DigineticaPrecision@K, MRR@K
SHAREYooChoose, DigineticaMRR@K, Hits@K
HGNNAYoochoose, DigineticaPrecision@K, MRR@K
MHCNLastFM, Douban, YelpPrecision@K, Recall@K, NDCG@K
HASRELastFM, Douban, YelpPrecision@K, Recall@K, NDCG@K
SHGCNBeidian, BeibeiRecall@K, NDCG@K
HHGRWeeplaces, CAMRa2011, DoubanRecall@K, NDCG@K
HyperTeNetAOTM, Goodreads, Spotify, ZhihuHR@K, NDCG@K, MAP@K
KHNNLast.FM, Book-Crossing, MovieLens-20MAUC, F1
), ArticleFig(id=1329455750906228958, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表10, caption=

应用于推荐系统的超图神经网络模型、数据集和评价指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
HyperRecAmazon, Etsy, GoodreadsHits@K, NDCG@K, MRR
H2SeqRecAMT, GoodreadsHR@K, NDCG@K
DHCFMovieLens-100K, CiteUlike-A, Baidu-Feed, Baidu-newsNDCG@K, Precision@K, Recall@K, Hits@K
SHCNAmazon, YelpHR, F1, NDCG
DHCNYoochoose, DigineticaPrecision@K, MRR@K
SHAREYooChoose, DigineticaMRR@K, Hits@K
HGNNAYoochoose, DigineticaPrecision@K, MRR@K
MHCNLastFM, Douban, YelpPrecision@K, Recall@K, NDCG@K
HASRELastFM, Douban, YelpPrecision@K, Recall@K, NDCG@K
SHGCNBeidian, BeibeiRecall@K, NDCG@K
HHGRWeeplaces, CAMRa2011, DoubanRecall@K, NDCG@K
HyperTeNetAOTM, Goodreads, Spotify, ZhihuHR@K, NDCG@K, MAP@K
KHNNLast.FM, Book-Crossing, MovieLens-20MAUC, F1
), ArticleFig(id=1329455751002697951, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 11, caption=

Common Datasets of Visual Object Classification

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集对象个数训练节点个数测试节点个数类别个数
ModelNet40123119843246840
NTU2 012163937367
), ArticleFig(id=1329455751082389728, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表11, caption=

视觉对象分类常用数据集

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集对象个数训练节点个数测试节点个数类别个数
ModelNet40123119843246840
NTU2 012163937367
), ArticleFig(id=1329455751178858721, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 12, caption=

Results of Different Models for Visual Object Classification (Accuracy) %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集
NTUModelNet40
GCN76.194.4
HGNN84.296.7
HGATs85.597.1
DeepHGNN85.697.2
MHGNNb85.597.5
HGWNN91.397.8
ResMultiHGNN92.198.7
), ArticleFig(id=1329455751258550498, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表12, caption=

视觉对象分类不同模型的结果(Accuracy)

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集
NTUModelNet40
GCN76.194.4
HGNN84.296.7
HGATs85.597.1
DeepHGNN85.697.2
MHGNNb85.597.5
HGWNN91.397.8
ResMultiHGNN92.198.7
), ArticleFig(id=1329455751321465059, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 13, caption=

Hypergraph Neural Network Models, Datasets and Evaluations Applied for Computer Vision

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
文献[138]CelebA-HQ, Paris Street View, Facades, Places2PSNR, SSIM, FID, L1 loss, L2 loss
HGCNN3DMAD, FA3DHTER, ACER, ACC, APCER, BPCER
S2HCNIndain Pines, Kennedy Space Center(KSC)Accuracy, Kappa coefficient
MGHMARS, iLIDS-VID, PRID-2011CMC, MAP
HGNNModelNet40, NTUAccuracy
HGATsModelNet40, NTUAccuracy
HGWNNModelNet40, NTUAccuracy
ResMultiHGNNModelNet40, NTUAccuracy
MHGNNbModelNet40, NTU, FacebookMAP , Accuracy
HANsGQA, VQA v2Accuracy
文献[140]VIOLIN, VLEP, VisualCOMET, MSVD, VISTAccuracy, BLEU-4, CIDEr, METEOR
SD-HNNHuman3.6M, MPI-INF-3DHPMPJPE,PCK, AUC
Hyper-GNNNTU RGB+D, Kinetics-SkeletonCS Accuracy, top-1 Accuracy, top-5 Accuracy
HI-GCNSUN, MS-COCO, ESP-GAMEMAP, CP, CR, CF1, OP, OR, OF1
AdaHGNNMS-COCO, NUS-Wide, Visual Genome, Pascal VOC 2007MAP, CF1, OF1
), ArticleFig(id=1329455751422128356, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表13, caption=

应用于计算机视觉的超图神经网络模型、数据集和评价

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
文献[138]CelebA-HQ, Paris Street View, Facades, Places2PSNR, SSIM, FID, L1 loss, L2 loss
HGCNN3DMAD, FA3DHTER, ACER, ACC, APCER, BPCER
S2HCNIndain Pines, Kennedy Space Center(KSC)Accuracy, Kappa coefficient
MGHMARS, iLIDS-VID, PRID-2011CMC, MAP
HGNNModelNet40, NTUAccuracy
HGATsModelNet40, NTUAccuracy
HGWNNModelNet40, NTUAccuracy
ResMultiHGNNModelNet40, NTUAccuracy
MHGNNbModelNet40, NTU, FacebookMAP , Accuracy
HANsGQA, VQA v2Accuracy
文献[140]VIOLIN, VLEP, VisualCOMET, MSVD, VISTAccuracy, BLEU-4, CIDEr, METEOR
SD-HNNHuman3.6M, MPI-INF-3DHPMPJPE,PCK, AUC
Hyper-GNNNTU RGB+D, Kinetics-SkeletonCS Accuracy, top-1 Accuracy, top-5 Accuracy
HI-GCNSUN, MS-COCO, ESP-GAMEMAP, CP, CR, CF1, OP, OR, OF1
AdaHGNNMS-COCO, NUS-Wide, Visual Genome, Pascal VOC 2007MAP, CF1, OF1
), ArticleFig(id=1329455751501820133, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 14, caption=

Common Datasets of Knowledge Base Question Answering

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集问题个数知识图谱
三元组个数实体个数关系个数
MetaQA4×1051.35×1054.3×1049
WebQuestionsSP4.7×1035.7×1061.8×106
), ArticleFig(id=1329455751568928998, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表14, caption=

知识库问答常用数据集

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集问题个数知识图谱
三元组个数实体个数关系个数
MetaQA4×1051.35×1054.3×1049
WebQuestionsSP4.7×1035.7×1061.8×106
), ArticleFig(id=1329455751665397991, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 15, caption=

Results of Different Models for Knowledge Base Question Answering(Hits@1) %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集
MetaQAWebQuestionsSP
KVMN96.246.7
GraftNet97.066.8
2HR-DR98.867.0
RecHyperNet99.168.4
), ArticleFig(id=1329455751740895464, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表15, caption=

知识库问答不同模型的结果(Hits@1)

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集
MetaQAWebQuestionsSP
KVMN96.246.7
GraftNet97.066.8
2HR-DR98.867.0
RecHyperNet99.168.4
), ArticleFig(id=1329455751803810025, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 16, caption=

Hypergraph Neural NetWork Models, Datasets and Evaluations Applied for Natural Language Processing

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
2HR-DRMetaQA, PQLHits@1, F1
RecHyperNetMetaQA, WebQuestionsSPHits@1
文献[144]WebQuestionsSPHits@1, F1
HSimplE,HypEJF17K, FB-AUTO, M-FB15KHits@K, MRR
HyperGAT20-Newsgroups, Reuters, Ohsumed, MovieReviewAccuracy
), ArticleFig(id=1329455751900279018, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表16, caption=

应用于自然语言处理的超图神经网络模型、数据集和评价指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
2HR-DRMetaQA, PQLHits@1, F1
RecHyperNetMetaQA, WebQuestionsSPHits@1
文献[144]WebQuestionsSPHits@1, F1
HSimplE,HypEJF17K, FB-AUTO, M-FB15KHits@K, MRR
HyperGAT20-Newsgroups, Reuters, Ohsumed, MovieReviewAccuracy
), ArticleFig(id=1329455751975776491, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 17, caption=

Hypergraph Neural Network Models, Datasets and Evaluation Applied for Stock Trend Prediction

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
STHGCNPrice data, Sector-Industry dataF1, SR
STHAN-SRNASDAQ, NYSE, TSESR, IRR, NDCG
HGTANThe data of stocks from ChinaAccuracy, Precision, Recall,
F1,MDD, IRR, SR
), ArticleFig(id=1329455752042885356, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表17, caption=

应用于预测股票趋势的超图神经网络模型、数据集和评价指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
STHGCNPrice data, Sector-Industry dataF1, SR
STHAN-SRNASDAQ, NYSE, TSESR, IRR, NDCG
HGTANThe data of stocks from ChinaAccuracy, Precision, Recall,
F1,MDD, IRR, SR
), ArticleFig(id=1329455752122577133, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 18, caption=

Hypergraph Neural Network Models, Datasets and Evaluation Applied for Chemistry and Medical

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
NHP-UiJO1366, iAF1260b, USPTOAUC, Recall@K
HCAEa subset of ADNI GOAccuracy
文献[149]ABIDEAccuracy
Hyper-SAGNNGPS, MovieLens, drug, wordnetscHi-CAUC, Micro-F, Macro-F1, ARI, AUPR
HUNeta subset of ADNI GO, ABIDEAccuracy, Sensitivity, Specificity
HGGANThe 219 subjects from ADNI databaseAccuracy, Sensitivity, Specificity
MRL-AHFThe 300 subjects from ADNI databaseAccuracy, Sensitivity, Specificity, AUC
), ArticleFig(id=1329455752189685998, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表18, caption=

应用于化学和医疗的超图神经网络模型、数据集和评价指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
NHP-UiJO1366, iAF1260b, USPTOAUC, Recall@K
HCAEa subset of ADNI GOAccuracy
文献[149]ABIDEAccuracy
Hyper-SAGNNGPS, MovieLens, drug, wordnetscHi-CAUC, Micro-F, Macro-F1, ARI, AUPR
HUNeta subset of ADNI GO, ABIDEAccuracy, Sensitivity, Specificity
HGGANThe 219 subjects from ADNI databaseAccuracy, Sensitivity, Specificity
MRL-AHFThe 300 subjects from ADNI databaseAccuracy, Sensitivity, Specificity, AUC
), ArticleFig(id=1329455752277766383, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=EN, label=Table 19, caption=

Hypergraph Neural Network Models, Datasets and Evaluation Applied for Traffic Forecast

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
GST-HCNCaltrans Performance Measurement
System Dataset
MAPE, MAE, RMSE
DSTHGCNBJMF1, HZMF19MAPE, MAE
DHATMETR-LA, PEMS-BAY, PEMS-D4,
PEMS-D8
MAE, RMSE, MAPE
), ArticleFig(id=1329455752340680944, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, language=CN, label=表19, caption=

应用于交通预测的超图神经网络模型、数据集和评价指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模型数据集评价指标
GST-HCNCaltrans Performance Measurement
System Dataset
MAPE, MAE, RMSE
DSTHGCNBJMF1, HZMF19MAPE, MAE
DHATMETR-LA, PEMS-BAY, PEMS-D4,
PEMS-D8
MAE, RMSE, MAPE
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202220483, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202220483, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202220483, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202220483, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1706803200000, fullTextJson=null, articleText=null, reference=null)
收藏切换
超图神经网络综述
林晶晶 1, 2, 3 , 冶忠林 1, 2 , 赵海兴 1, 2, * , 李卓然 1, 2
计算机研究与发展 | 人工智能和模式识别 2024,61(2): 362-384
收起
收藏切换
计算机研究与发展 |人工智能和模式识别 2024 , 61 (2) : 362 -384
超图神经网络综述
全屏
[Author(id=1329455743402619041, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=ljj_mail@126.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329455743478116517, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743402619041, language=EN, stringName=Jingjing Lin, firstName=Jingjing, middleName=null, lastName=Lin, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 3, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008
3Department of Information Engineering, Xining Urban Vocational & Technical College, Xining 810003, bio={"content":"

bron in 1986. PhD candidate, lecturer. Her main research interests include graph neural networks and hypergraph neural networks

"}, bioImg=null, bioContent=

bron in 1986. PhD candidate, lecturer. Her main research interests include graph neural networks and hypergraph neural networks

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455743830438054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743402619041, language=CN, stringName=林晶晶, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 3, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008
3西宁城市职业技术学院信息工程系 西宁 810003, bio={"img":"G5rQbtrDWA86ZLRSo9TXJg==","content":"

1986年生. 博士研究生,讲师. 主要研究方向为图神经网络、超图神经网络

"}, bioImg=G5rQbtrDWA86ZLRSo9TXJg==, bioContent=

1986年生. 博士研究生,讲师. 主要研究方向为图神经网络、超图神经网络

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)]), AuthorCompany(id=1329455743322927261, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=3, ext=[AuthorCompanyExt(id=1329455743327121566, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743322927261, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Department of Information Engineering, Xining Urban Vocational & Technical College, Xining 810003), AuthorCompanyExt(id=1329455743335510175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743322927261, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3西宁城市职业技术学院信息工程系 西宁 810003)])]), Author(id=1329455743893352616, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329455743989821611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743893352616, language=EN, stringName=Zhonglin Ye, firstName=Zhonglin, middleName=null, lastName=Ye, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008, bio={"content":"

born in 1989. PhD, associate professorr, PhD supervisor. Member of CCF. His main research interests include graph neural networks, knowledge extraction, and network representation learning

"}, bioImg=null, bioContent=

born in 1989. PhD, associate professorr, PhD supervisor. Member of CCF. His main research interests include graph neural networks, knowledge extraction, and network representation learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455744086290604, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455743893352616, language=CN, stringName=冶忠林, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008, bio={"img":"VVg0HWqw4/9zUsqlvaF/EA==","content":"

1989年生. 博士,副教授,博士生导师. CCF会员. 主要研究方向为图神经网络、知识抽取、网络表示学习

"}, bioImg=VVg0HWqw4/9zUsqlvaF/EA==, bioContent=

1989年生. 博士,副教授,博士生导师. CCF会员. 主要研究方向为图神经网络、知识抽取、网络表示学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)])]), Author(id=1329455744165982382, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=h.x.zhao@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329455744241479857, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744165982382, language=EN, stringName=Haixing Zhao, firstName=Haixing, middleName=null, lastName=Zhao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008, bio={"content":"

born in 1969. PhD, professor, PhD supervisor. Member of CCF. His main research interests include complex network, graph neural networks, machine translation, hypergraph theory, and network reliability

"}, bioImg=null, bioContent=

born in 1969. PhD, professor, PhD supervisor. Member of CCF. His main research interests include complex network, graph neural networks, machine translation, hypergraph theory, and network reliability

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455744312783026, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744165982382, language=CN, stringName=赵海兴, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008, bio={"img":"pS/TFCPeGGYENBVzrfBeOA==","content":"

1969年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为复杂网络、图神经网络、机器翻译、超图理论、网络可靠性

"}, bioImg=pS/TFCPeGGYENBVzrfBeOA==, bioContent=

1969年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为复杂网络、图神经网络、机器翻译、超图理论、网络可靠性

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)])]), Author(id=1329455744400863412, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329455744480555191, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744400863412, language=EN, stringName=Zhuoran Li, firstName=Zhuoran, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1College of Computer, Qinghai Normal University, Xining 810008
2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008, bio={"content":"

born in 1996. Master. His main research interests include data mining and graph neural networks

"}, bioImg=null, bioContent=

born in 1996. Master. His main research interests include data mining and graph neural networks

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329455744543469752, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, authorId=1329455744400863412, language=CN, stringName=李卓然, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1青海师范大学计算机学院 西宁 810008
2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008, bio={"img":"cMiNtBIHDxwc9c2wbVX1Bg==","content":"

1996年生. 硕士. 主要研究方向为数据挖掘、图神经网络

"}, bioImg=cMiNtBIHDxwc9c2wbVX1Bg==, bioContent=

1996年生. 硕士. 主要研究方向为数据挖掘、图神经网络

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329455743150960791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=1, ext=[AuthorCompanyExt(id=1329455743159349400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Computer, Qinghai Normal University, Xining 810008), AuthorCompanyExt(id=1329455743163543705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743150960791, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1青海师范大学计算机学院 西宁 810008)]), AuthorCompany(id=1329455743243235482, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, xref=2, ext=[AuthorCompanyExt(id=1329455743255818395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008), AuthorCompanyExt(id=1329455743260012700, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329455739829072004, companyId=1329455743243235482, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008)])])]
林晶晶1, 2, 3 , 冶忠林1, 2, 赵海兴1, 2, * , 李卓然1, 2
作者信息
  • 1青海师范大学计算机学院 西宁 810008
  • 2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008
  • 3西宁城市职业技术学院信息工程系 西宁 810003
通讯作者:
赵海兴()
作者简介:

1986年生. 博士研究生,讲师. 主要研究方向为图神经网络、超图神经网络

1989年生. 博士,副教授,博士生导师. CCF会员. 主要研究方向为图神经网络、知识抽取、网络表示学习

1969年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为复杂网络、图神经网络、机器翻译、超图理论、网络可靠性

1996年生. 硕士. 主要研究方向为数据挖掘、图神经网络

Survey on Hypergraph Neural Networks
Jingjing Lin1, 2, 3 , Zhonglin Ye1, 2, Haixing Zhao1, 2, * , Zhuoran Li1, 2
Affiliations
  • 1College of Computer, Qinghai Normal University, Xining 810008
  • 2State Key Laboratory of Tibetan Intelligent Information Processing and Application (Qinghai Normal University), Xining 810008
  • 3Department of Information Engineering, Xining Urban Vocational & Technical College, Xining 810003
  • bron in 1986. PhD candidate, lecturer. Her main research interests include graph neural networks and hypergraph neural networks

    born in 1989. PhD, associate professorr, PhD supervisor. Member of CCF. His main research interests include graph neural networks, knowledge extraction, and network representation learning

    born in 1969. PhD, professor, PhD supervisor. Member of CCF. His main research interests include complex network, graph neural networks, machine translation, hypergraph theory, and network reliability

    born in 1996. Master. His main research interests include data mining and graph neural networks

出版时间: 2024-02-02 doi: 10.7544/issn1000-1239.202220483
文章导航
收藏切换

近年来,图神经网络借助大量数据和超强计算能力在推荐系统和自然语言处理等应用领域取得显著成效,它主要处理具有成对关系的图数据. 但许多现实网络中的对象之间的关系是复杂的非成对关系,如科研合作网络、蛋白质网络等. 若直接用图结构将这种复杂的关系表示为成对关系,会导致信息丢失. 超图是一种灵活的建模工具,可以展现出图无法完整刻画的高阶关系,弥补了图的不足. 鉴于此,研究者开始关心如何在超图上设计神经网络,并相继提出应用于下游任务的超图神经网络模型(hypergraph neural network,HGNNs). 故对现有的超图神经网络模型进行综述,首先全面回顾超图神经网络在过去3年的研究历程;其次根据设计超图神经网络采用的方法不同对其进行分类,并详细地阐述代表性的模型;然后介绍了超图神经网络的应用领域;最后总结和探讨了超图神经网络未来的研究方向.

超图  /  超图神经网络  /  分类  /  图  /  图神经网络

In recent years, graph neural networks have achieved remarkable results in application fields such as recommendation systems and natural language processing with the help of large amounts of data and supercomputing power, and they mainly deal with graph data with pairwise relationships. However, in many real-world networks, the relationships between objects are more complex and beyond pairwise, such as scientific collaboration networks, protein networks, and others. If we directly use a graph to represent the complex relationships as pairwise relations, which will lead to a loss of information. Hypergraph is a flexible modeling tool, which shows higher-order relationships that cannot be fully described by a graph, making up for the shortage of graph. In light of this, scholars begin to care about how to develop neural networks on hypergraph, and successively put forward many hypergraph neural network models. Therefore, we overview the existing hypergraph neural network models. Firstly, we comprehensively review the development of the hypergraph neural network in the past three years. Secondly, we propose a new classification method according to the design method of hypergraph neural networks, and elaborate on representative models. Then, we introduce the application areas of hypergraph neural networks. Finally, the future research direction of hypergraph neural networks are summarized and discussed.

hypergraph  /  hypergraph neural network  /  classification  /  graph  /  graph neural networks
林晶晶, 冶忠林, 赵海兴, 李卓然. 超图神经网络综述. 计算机研究与发展, 2024 , 61 (2) : 362 -384 . DOI: 10.7544/issn1000-1239.202220483
Jingjing Lin, Zhonglin Ye, Haixing Zhao, Zhuoran Li. Survey on Hypergraph Neural Networks[J]. Journal of Computer Research and Development, 2024 , 61 (2) : 362 -384 . DOI: 10.7544/issn1000-1239.202220483
卷积神经网络[1](convolutional neural network, CNN)、循环神经网络[2](recurrent neural network, RNN)等传统的深度学习模型在图像、音频等任务中取得较好的效果,这取决于它们被具有平移不变性和局部连通性的欧氏数据来表示.CNN的核心思想局部连接、参数共享、池化采样和多层使用[3],降低了网络模型的复杂度,减少了权值的数量,却未减弱其表达能力.RNN拥有处理可变长度的序列数据的能力,能够挖掘数据中的时序信息和语义信息. 长短期记忆[4](long short-term memory,LSTM)是一种采用门控机制改进的RNN,而门控递归单元[5](gated recurrent unit,GRU)是LSTM的变体,只包含重置门和更新门,简化了LSTM的架构. 尽管这些传统的深度学习方法能够有效地提取欧氏数据的特征,但许多实际应用场景中还存在一种用图结构表示的非欧氏数据,如社交网络、交通网络、知识图谱和蛋白质网络等[6-8]. 因此,研究者对深度学习方法在图数据上的扩展越来越感兴趣.
近年来,受CNN和RNN的启发,研究者们创新性地设计了用于处理图数据的深度学习方法,统称为图神经网络(graph neural networks,GNNs). 由于GNNs具有丰富的表达能力、灵活的建模能力和端到端的训练能力,被广泛地应用于图分析领域,突破性地优化和提高了节点分类[9-11]、预测[12-13]、视觉分析[14]和自然语言处理[15-16]等任务的性能. 在文献[17]中将GNNs分为循环图神经网络、卷积图神经网络、图自动编码器和时空图神经网络. 其中,卷积图神经网络也可称为图卷积神经网络(graph convolution networks,GCNs)是最流行和发展最快的图神经网络之一. 与CNN相比,GCNs需要在图上设计卷积操作来刻画节点的邻域结构. 依据卷积操作定义方式的不同,GCNs分为谱域方法和空域方法. 谱域方法通过卷积定理在谱域定义卷积操作; 空域方法通过聚合函数汇集每个顶点及其邻域节点的方式定义卷积操作.
在现实世界的网络中,虽然图可以很好地刻画对象间的成对关系,但是对象间还存在大量比成对关系更加复杂的非成对关系. 如果简单地用图表示对象间的复杂高阶关系,可能会导致信息丢失或信息冗余. 例如,多位作者撰写了一篇文章,每位作者视为顶点,若这几位作者中的任何2位都非必要直接联系[18],用图直接将他们两两连接,会产生信息冗余,无法恰当地描述出多人撰写文章的合著关系,使得展示的信息不准确. 超图由顶点集和超边集组成,每条超边可以连接多个顶点而非仅2个顶点. 故超图拥有描述顶点间复杂高阶关系的能力,可用于建模具有高阶相互作用的复杂网络和系统. 因此,若用一条超边连接多位合著者,能够更加简洁和准确地刻画出多位作者的合著关系.
近3年来,受GNNs和超图建模优势的启发,研究者们开始关注如何将GNNs扩展到超图,设计出基于超图的神经网络,改善具体应用的性能.2019年,Feng等人[19]提出第1个超图神经网络 (hypergraph neural networks,HGNN),将GCNs的谱方法很自然地扩展到超图上,并设计了超图卷积操作. 同年,Yadati等人[20]设计HyperGCN实现在超图上处理半监督分类问题. 随后,陆续出现的大量超图神经网络模型被充分运地用到计算机视觉[21-22]、推荐系统[23-24]、生物化学[25-26]等领域并取得显著的成绩. 超图神经网络已逐渐成为一个新的研究热点. 目前,对于上文中提到的CNN、GNNs和GCNs,存在多篇经典的综述性文献对它们进行全面的归纳和总结. 例如,文献[17]提供了一种划分GNNs的策略,总结了GNNs的数据集、开源代码和模型评估等. 在文献[27]中详细地介绍了GNNs模型及其变体,从结构场景、非结构场景和其他场景详细地概括了GNNs的应用;同时探讨了4个开放性问题. 文献[28]系统地总结了CNN的发展史、网络结构、训练方法及应用场景等. 文献[6]全面梳理了GCNs的谱域方法和空域方法、最新进展、相关应用和发展趋势. 除此之外,还有大量的文献[29-31]从非欧空间、具体应用领域等方面总结和展望了GNNs.
Gao等人[32]详细介绍了超图学习方法,但关于超图神经网络的分类方法、模型分析和应用领域等方面的探讨和总结仍为空白. 因此,系统地综述现有超图神经网络模型十分必要,有助于推动超图神经网络今后在更多的领域中发展和应用. 本文全面地梳理了超图神经网络近3年的发展历程、代表性模型及其未来研究方向. 根据设计模型采用的方法不同,将超图神经网络分为超图卷积神经网络、超图循环神经网络和超图生成对抗网络3类,并根据实现卷积的方法细分超图卷积神经网络. 重点阐述了每类中的代表性模型,并做分析和比较. 同时讨论了超图神经网络未来的研究工作.
在过去的几年里,GNNs作为一种强大且实用的工具,推动深度学习在图数据上快速发展. 大量研究表明GNNs在计算机视觉[33-35]、自然语言处理[36-37]、交通网络[38-40]、推荐系统[41-43]、生物化学[44-46]、知识图谱[47-49]等各领域获得成功.GNNs主要处理图数据,但现实网络中还存在超越二元的关系,若用图简单地表示,会丢失对象之间潜在的高阶关系. 超图[50]是一种建模复杂关系的灵活工具,其包含顶点集和超边集,利用每条超边连接多个顶点表示潜在的高阶关系. 本质上,每条超边是顶点的子集,即一条超边连接多个顶点,而非只连接2个顶点. 因此,超图作为图的一种推广,可以刻画出更加丰富的信息,充分地挖掘图中顶点之间潜在的高阶交互.
在过去3年中,超图神经网络吸引了很多研究者探究如何在超图上设计神经网络,以及如何将其应用于各个领域. 借鉴GNNs中的卷积运算、注意力机制、生成对抗等技术,超图神经网络迅速发展. 研究者们在超图上开发出超图卷积、超图注意力和超图生成对抗网络等,相继地涌现出大量的超图神经网络模型.图1详细地展示了从2019—2022年初超图神经网络模型发展进程的时序情况,横轴表示时间(以年为间隔),纵轴表示模型,着重列出了每个模型的简称和作者信息.
从图1中可以清晰地看出超图神经网络模型的数量逐年递增,研究重心也发生了变化,具体如下所述. 首先,超图神经网络建模的超图结构从无向超图扩展到有向超图,以及从静态超图结构拓展到动态超图结构. 分别设计出有向超图神经网络,如有向超图卷积网络 (directed hypergraph convolutional networks,DHGCN)[62]和有向超图神经网络(directed hypergraph neural network,DHN)[104];以及动态超图神经网络,如DHGNNa(dynamic hypergraph neural network)[52]、自适应超图神经网络(adaptive hypergraph neural network,AdaHGNN)[71]和自适应超图卷积网络(adaptive hypergraph convolutional network,AHGCN) [94]. 其次,超图神经网络的消息传递方式从单纯的超图卷积转变到引入注意力机制、循环神经网络和生成对抗网络,再通过组合它们,设计出基于注意力机制的超图神经网络,如Hyper-SAGN[25]、超图注意力网络HANs[60]和HyperGAT[75];超图卷积循环神经网络(hypergraph convolutional recurrent neural network,HGC-RNN)[66];超图生成对抗网络 (hypergraph generative adversarial network,HGGAN)[101]和多模态表示学习与对抗超图融合(multimodal representation learning and adversarial hypergraph fusion,MRL-AHF)[102]框架. 再次,超图神经网络的维度从单通道超图卷积升级到多通道,设计出双通道和多通道超图神经网络,如双通道超图协同滤波(dual channel hypergraph collaborative filtering,DHCF)[67]、双通道超图卷积网络 (dual channel hypergraph convolutional network,DHCN)[62]和多通道超图卷积网络 (multi-channel hypergraph convolutional network, MHCN)[83]. 最后,超图神经网络的应用从最初的引文网络分类任务拓展到推荐系统、自然语言处理、股票和交通预测等任务,设计出多种适用于各领域下游任务的超图神经网络,如信号超图卷积网络 (signed hypergraph convolutional network,SHCN)[24]、2HR-DR[62]、时空超图卷积网络(spatio-temporal hypergraph convolution network,STHGCN)[76]和地理-语义-时间超图卷积网络(geographic-semantic-temporal hypergraph convolutional network,GST-HCN)[55].
本节给出了涉及的定义及常见符号的含义,有助于理解本文中的公式. 下面依次给出符号描述,以及图、超图和有向超图的定义.
本文中用到的符号及其含义如表1所示.
$G = ({V_G},{E_G},{{\boldsymbol{W}}_G})$是一个图,其中$ {V_G} $是顶点的集合,$ {E_G} $是边的集合.$ \left| {{V_G}} \right| = {N_G} $和$ \left| {{E_G}} \right| = {M_G} $,$ {N_G} \in \mathbb{R} $,$ {M_G} \in \mathbb{R} $分别表示图中顶点和边的数目. 若令$ v_G^i $和$ v_G^j $是$G$中的顶点,则${e_{ij}} = (v_G^i,v_G^j) \in {E_G}$代表$G$中连接$ v_G^i $和$ v_G^j $的一条边.${{\boldsymbol{W}}_G}$是边的权重矩阵. 图的邻接矩阵${\boldsymbol{A}} \in {\mathbb{R}^{{N_G} \times {N_G}}}$定义为
$ {a}_{ij}=\left\{\begin{split}1,\quad {\mathrm{if}}\;{e}_{ij}\in {E}_{G},\\ 0,\quad \mathrm{if}\;{e}_{ij}\notin {E}_{G},\end{split}\right. $
其中${a_{ij}}$表示${\boldsymbol{A}}$中的元素,$i,j = 1,2,…,{N_G}$.
$\mathcal{G} = ({\mathcal{V}_\mathcal{G}},{\mathcal{E}_\mathcal{G}},{\mathcal{W}_\mathcal{G}})$代表一个超图,${\mathcal{V}_\mathcal{G}}$是顶点集,${\mathcal{E}_\mathcal{G}}$是超边集.$\mathcal{G}$包含$|{\mathcal{V}_\mathcal{G}}| = {N_\mathcal{G}}$个顶点和$ |{\mathcal{E}_\mathcal{G}}| = {M_\mathcal{G}} $条超边. 若令$ v_\mathcal{G}^i $是顶点,则超边$e_\mathcal{G}^i = \{ v_\mathcal{G}^{{m_{\text{e}}}},…,v_\mathcal{G}^{{n_{\text{e}}}}\} $,其中$1 \leqslant {m_{\text{e}}} \leqslant {N_\mathcal{G}}$,$1 \leqslant {n_{\text{e}}} \leqslant {N_\mathcal{G}}$.${\mathcal{W}_\mathcal{G}}$表示超边的权重矩阵.
通常情况下,超图用关联矩阵${{\boldsymbol{H}}_\mathcal{G}} \in {\mathbb{R}^{{N_\mathcal{G}} \times {M_\mathcal{G}}}}$表示,定义为
$ {h}_{ij}=h({v}_{\mathcal{G}}^{i},{e}_{\mathcal{G}}^{j})=\left\{\begin{array}{c}1, \quad \mathrm{if}\; {v}_{\mathcal{G}}^{i}\in {e}_{\mathcal{G}}^{j},\\ 0,\quad\mathrm{if}\; {v}_{\mathcal{G}}^{i}\notin {e}_{\mathcal{G}}^{j},\end{array}\right. $
其中${h_{ij}}$是${{\boldsymbol{H}}_\mathcal{G}}$的元素,$i = 1,2,…,{N_\mathcal{G}}$,$j = 1,2,…,{M_\mathcal{G}}$.
若$ v_\mathcal{G}^i \in {\mathcal{V}_\mathcal{G}} $,节点的度指包含$ v_\mathcal{G}^i $的超边的数目,记为
$ d(v_\mathcal{G}^i) = \sum\limits_{{e_\mathcal{G}} \in {\mathcal{E}_\mathcal{G}}} {h(v_\mathcal{G}^i,{e_\mathcal{G}}){\mathcal{W}_\mathcal{G}}({e_\mathcal{G}})} . $
若$ e_{\mathcal{G}}^j\in\mathcal{E}_{\mathcal{G}} $,超边的度指超边$ e_\mathcal{G}^j $中包含的顶点的数目,记为
$ d(e_\mathcal{G}^j) = \sum\limits_{v_\mathcal{G}^i \in {\mathcal{V}_\mathcal{G}}} {h(v_\mathcal{G}^i,e_\mathcal{G}^j)} . $
$ {{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}} $和$ {{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}} $分别表示顶点和超边的度矩阵. 有关超图的详细说明可参见文献[50].
事实上,图是超图的一个特例,当超图中的每条超边只包含2个顶点时,超图就退化为一个普通图. 如图2(a)是一个普通图,每条边连接2个顶点;图2(b)是一个超图,每条超边连接3个或4个顶点. 当图2(b)中的每条超边只连接2个顶点时,超图就退化为一个图,即图2(c)(d).
$ {\mathcal{G}_{\mathrm{D}}} = ({\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}},{\mathcal{E}_{{\mathcal{G}_{\mathrm{D}}}}},{\mathcal{W}_{{\mathcal{G}_{\mathrm{D}}}}}) $代表有向超图,${\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}}$是顶点集,${\mathcal{E}_{{\mathcal{G}_{\mathrm{D}}}}}$是超弧集,${\mathcal{W}_{{\mathcal{G}_{\mathrm{D}}}}}$表示超弧的权重矩阵. $|{\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}}| = {N_{{\mathcal{G}_{\mathrm{D}}}}}$和$ |\mathcal{E}_{\mathcal{G}_{\mathrm{D}}}|=\mathit{M}_{\mathcal{G}_{\mathrm{D}}} $分别表示顶点和超弧的数目. 每条超弧 $ {\vec{e}}_{i}=({\vec{e}}_{i}^{\text{ tail}},{\vec{e}}_{i}^{\text{ head}})\in {\mathcal{E}_{{\mathcal{G}_{\mathrm{D}}}}} $,其中$ \vec e_i^{\;{\text{head}}} \in {\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}} $表示超弧${\vec e_i}$的弧头,$\vec e_i^{\;{\text{tail}}} \in {\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}}$表示超弧${\vec e_i}$的弧尾. 且$\vec e_i^{\;{\text{tail}}} \cap \vec e_i^{\;{\text{head}}} = \varnothing $,$ \vec e_i^{\;{\text{tail}}} \ne \varnothing $,$\vec e_i^{\;{\text{head}}} \ne \varnothing $. 关于有向超图的其他描述可参见文献[50,57].
在本节中,梳理出一种超图神经网络的分类,依据设计模型采用的方法不同,将现有超图神经网络模型划分为超图卷积神经网络、超图循环神经网络和超图生成对抗网络. 进一步将超图卷积神经网络细分为3类,详细地介绍每类中的代表性模型.
超图卷积神经网络本质上是将GCNs显式或隐式地扩展得到超图上,又或者引入注意力机制,设计出适用于超图上的卷积操作. 根据设计卷积算子的策略不同,将超图卷积神经网络进一步细分为基于谱域的超图卷积神经网络、基于普通图的超图神经网络和基于注意力机制的超图神经网络.
基于谱域的超图卷积神经网络是利用图信号和超图谱理论,巧妙地借助傅里叶变换和小波变换,将GCNs隐式地扩展到超图上. 本节根据超图结构是否有向、是否动态将基于谱域的超图卷积神经网络细分为无向超图神经网络、有向超图神经网络和动态超图神经网络.
首先回顾一下谱域的GCN[114],可以帮助读者更好地理解基于谱域的超图卷积神经网络的演化. Spectral CNN[115]最早利用卷积定理在图上设计谱卷积算子,但它的计算复杂度高. 为解决这一问题,ChebyNet(ChebyShev network)[116]采用$K$阶切比雪夫多项式近似卷积核,此时图卷积算子仅依赖于顶点的$K$阶邻居. Kipf 等人[114]继续简化参数,提出只考虑1阶邻居的GCN,提升了图上半监督学习的性能. GCN的卷积层定义为:
$ {{\boldsymbol{H}}}^{(l+1)}=\sigma ({\widetilde {\boldsymbol{D}}^{ - {{{}^1/{}_2}}}}\widetilde {\boldsymbol{A}}{\widetilde {\boldsymbol{D}}^{ - {{{}^1/{}_2}}}}{{\boldsymbol{H}}}^{(l)}{{\boldsymbol{\varTheta}} }_{G}^{(l)}), $
其中$\widetilde{\boldsymbol{ A}} = {\boldsymbol{A}} + {\boldsymbol{I}}$,$\widetilde {\boldsymbol{D}}$代表$\widetilde{\boldsymbol{ A}}$的度矩阵. 若令 ${{\boldsymbol{S}}_G} = {\widetilde{\boldsymbol{ D}}^{ - {{{}^1/{}_2}}}}\widetilde{\boldsymbol{ A}}{\widetilde{\boldsymbol{ D}}^{ - {{{}^1/{}_2}}}}$,则式(5)可以重写为
$ {{\boldsymbol{H}}}^{(l+1)}=\sigma ({{\boldsymbol{S}}}_{G}{{\boldsymbol{H}}}^{(l)}{{\boldsymbol{\varTheta}} }_{G}^{(l)}), $
其中${{\boldsymbol{H}}^{(l)}}$和${{\boldsymbol{H}}^{(l + 1)}}$分别是第$l$和$l + 1$层顶点的隐层表示,${\boldsymbol{\varTheta}} _G^{(l)}$是参数矩阵.
通常情况下超图都是无向超图,即超边不具有方向性,本节主要介绍以无向超图为拓扑结构的超图的神经网络模型.
1)HGNN
利用超图建模数据间的高阶关系,充分地考虑了对象间的高阶关系和多模态特性. 为更好地理解HGNN定义的卷积,首先要理解超图的拉普拉斯矩阵. Zhou等人[117]定义了超图拉普拉斯矩阵,$ {\mathcal{L}_\mathcal{G}} = {\boldsymbol{I}} - {\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{}^1/{}_2}}{{\boldsymbol{H}}_\mathcal{G}}{\mathcal{W}_\mathcal{G}}{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{}^1/{}_2}} $. 其特征分解为$ {\mathcal{L}_\mathcal{G}} = {{\boldsymbol{U}}_\mathcal{G}}{\boldsymbol{\varLambda}} {\boldsymbol{U}}_\mathcal{G}^\mathrm{T} $,其中$ {\boldsymbol{\varLambda}} $是特征值的对角矩阵,$ {{\boldsymbol{U}}_\mathcal{G}} $是特征向量矩阵. 依据超图的拉普拉斯矩阵的定义,将GCN类比到超图上,设计出超图卷积算子.
对于超图上的信号$x$,首先将$x$经过傅里叶变换从空域转换到谱域,再在谱域上执行卷积操作,然后利用傅里叶逆变换将卷积结果从谱域转换到空域. 因此,通过傅里叶变换和逆变换,信号$x$与滤波器$g$的谱卷积可定义为
$ g\;{\ast }_{\mathcal{G}} \;x={{\boldsymbol{U}}}_{\mathcal{G}}(({{\boldsymbol{U}}}_{\mathcal{G}}^{\text{T}}g)\odot ({{\boldsymbol{U}}}_{\mathcal{G}}^{\text{T}}x))={{\boldsymbol{U}}}_{\mathcal{G}}g({\boldsymbol{\varLambda}} ){{\boldsymbol{U}}}_{\mathcal{G}}^{\text{T}}x, $
其中$ g({\boldsymbol{\varLambda}} ) $是傅里叶系数的函数.
由于傅里叶变换及其逆变换的计算成本高,采用截断切比雪夫多项式作为其近似多项式,谱卷积被重新定义为
$ g\;{\ast }_{\mathcal{G}}\;x\approx {\displaystyle \sum _{k=1}^{K}{\varTheta }_{k}{T}_{k}({\hat{\mathcal{L}}_\mathcal{G}})x}, $
其中${T_k}(y) = 2y{T_{(k - 1)}}(y) - {T_{(k - 2)}}(y)$代表切比雪夫多项式的递归形式,${T_{k}}({\hat {\mathcal{L}}}_{\mathcal{G}})$是变量为拉普拉斯算子$ {\hat{\mathcal{L}}}_{\mathcal{G}}=\dfrac{2}{{\lambda }_{\mathcal{G}\text{max}}}{\hat{\mathcal{L}}}_{\mathcal{G}}-{\boldsymbol{I}} $的截断切比雪夫多项式.
若令$K = 1$且$ {\lambda _{\mathcal{G}{\text{max}}}} = 2 $,降低阶数和减少计算的谱卷积定义为
$ g\;{ * }_{\mathcal{G}}\;x \approx {\varTheta }_{0}x-{\varTheta }_{1}{{\boldsymbol{D}}}_{{\mathcal{V}}_{\mathcal{G}}}^{ - {{}^1/{}_2}}{{\boldsymbol{H}}}_{\mathcal{G}}{\mathcal{W}}_{\mathcal{G}}{{\boldsymbol{D}}}_{{\mathcal{E}}_{\mathcal{G}}}^{-1}{{\boldsymbol{H}}}_{\mathcal{G}}^{\text{T}}{{{\boldsymbol{D}}}_{{\mathcal{V}}_{\mathcal{G}}}^{-{{}^1/{}_2}}}{x}, $
其中${\varTheta _0}$和${\varTheta _1}$是参数. 若用式(10)中的$\varTheta $表示${\varTheta _0}$和${\varTheta _1}$来进一步约束参数数量:
$ \left\{\begin{array}{l}{\varTheta }_{0}={ - {{}^1/{}_2}}\varTheta {{\boldsymbol{D}}}_{{\mathcal{V}}_{\mathcal{G}}}^{ - {{}^1/{}_2}}{{\boldsymbol{H}}}_{\mathcal{G}}{\mathcal{W}}_{\mathcal{G}}{{\boldsymbol{D}}}_{{\mathcal{E}}_{\mathcal{G}}}^{-1}{{\boldsymbol{H}}}_{\mathcal{G}}^{\text{T}}{{\boldsymbol{D}}}_{{\mathcal{V}}_{\mathcal{G}}}^{ - {{}^1/{}_2}},\\ {\varTheta }_{1}={ - {{}^1/{}_2}}\varTheta ,\end{array}\right. $
则式(10)代入式(9)得到1阶卷积为
$ {\text{ }}g{\text{ }}{ * _\mathcal{G}}{\text{ }}x \approx \varTheta {\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{H}}_\mathcal{G}}{\mathcal{W}_\mathcal{G}}{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ -{{{}^1/{}_2}}}}{x}. $
最终,采用矩阵形式表示HGNN的超图卷积层定义为
$ {{\boldsymbol{Q}}}_{\mathcal{G}}^{(l+1)}=\sigma ({{\boldsymbol{D}}}_{{\mathcal{V}}_{\mathcal{G}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{H}}}_{\mathcal{G}}{\mathcal{W}}_{\mathcal{G}}{{\boldsymbol{D}}}_{{{\mathcal{E}}}_{\mathcal{G}}}^{-1}{{\boldsymbol{H}}}_{\mathcal{G}}^{\text{T}}{{\boldsymbol{D}}}_{{\mathcal{V}}_{\mathcal{G}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{Q}}}_{\mathcal{G}}^{(l)}{{\boldsymbol{\varTheta}} }_{\mathcal{G}}^{(l)}), $
其中${\boldsymbol{Q}}_\mathcal{G}^{(l)}$和${\boldsymbol{Q}}_\mathcal{G}^{(l + 1)}$分别是第$l$和$l + 1$层顶点的隐层表示. 若令${{\boldsymbol{P}}_\mathcal{G}} = {\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{H}}_\mathcal{G}}{\mathcal{W}_\mathcal{G}}{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}$,则式(12)可重写为
$ {\boldsymbol{Q}}_\mathcal{G}^{(l + 1)} = \sigma ({{\boldsymbol{P}}_\mathcal{G}}{\boldsymbol{Q}}_\mathcal{G}^{(l)}{\boldsymbol{\varTheta}} _\mathcal{G}^{(l)}). $
从式(6)(13)可以发现GCN和HGNN的卷积计算形式很相似,都是矩阵相乘,只是两者描述数据关系所使用的建模工具不同,分别是图和超图,用${{\boldsymbol{S}}_G}$和${{\boldsymbol{P}}_\mathcal{G}}$表示. 与传统的超图学习方法相比,HGNN未使用超图拉普拉斯矩阵的逆运算,节约了计算成本.
2)行归一化超图卷积层
文献[118]给出另一种超图卷积的形式,即行归一化超图卷积层,定义为
$ {\boldsymbol{Q}}_\mathcal{G}^{(l + 1)} = \sigma ({\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - 1}{{\boldsymbol{H}}_\mathcal{G}}{\mathcal{W}_\mathcal{G}}{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{Q}}_\mathcal{G}^{(l)}{\boldsymbol{\varTheta}} _\mathcal{G}^{(l)}). $
3)MultiHGNN
Huang等人[93]提出多超图神经网络 (multi-hypergraph neural network,MultiHGNN)学习具有多模态特性的数据,用超图建模每种模态,分别在每个超图上执行卷积操作,采用均值方法生成顶点的最终表示;而HGNN在拼接所有模态对应的超图后,执行卷积操作得到顶点的最终表示.HGNN与MultiHGNN的具体框架比较如图3所示.
4)HplapGCN
Ma等人[119]将超图拉普拉斯${\mathcal{L}_\mathcal{G}}$推广到超图p-拉普拉斯$ {{\mathcal{L}}}_{{\mathrm{p}}\mathcal{G}} $. 鉴于超图p-拉普拉斯能够保存超图概率分布的几何特性,Fu等人[54]专门为超图p-拉普拉斯设计了超图p-拉普拉斯图卷积网络 (hypergraph p-Laplacian graph convolutional network,HpLapGCN). HpLapGCN的卷积层定义为
$ {{\boldsymbol{Q}}}_{{{\mathcal{L}}}_{\text{p}\mathcal{G}}}^{(l+1)}=\sigma ({{\boldsymbol{H}}}_{{{\mathcal{L}}}_{\text{p}\mathcal{G}}}{{\boldsymbol{Q}}}_{{{\mathcal{L}}}_{\text{p}\mathcal{G}}}^{(l)}{{\boldsymbol{\varTheta}} }_{{{\mathcal{L}}}_{\text{p}\mathcal{G}}}^{(l)}), $
其中$ {{\boldsymbol{H}}}_{{{\mathcal{L}}}_{{\mathrm{p}}\mathcal{G}}}=\dfrac{2}{{\lambda }_{{\mathcal{L}}_{{\mathrm{p}}\mathcal{G}}\mathrm{max}}}{\mathcal{L}}_{\text{p}\mathcal{G}}-{\boldsymbol{I}} $,$ \lambda_{L_{\mathrm{p}\mathcal{G}\mathrm{max}}} $是$ {{\mathcal{L}}}_{{\mathrm{p}}\mathcal{G}} $的最大特征值,关于$ {{\mathcal{L}}}_{{\mathrm{p}}\mathcal{G}} $的详细计算请查阅文献[68].$ {\varTheta }_{{{\mathcal{L}}}_{{\mathrm{p}}\mathcal{G}}}^{(l)} $是权重参数.
5)HWNN
为避免耗时的拉普拉斯分解,Sun等人[72]采用多项式近似的小波变换代替傅里叶变换,设计出适用于异构超图的超图小波神经网络(hypergraph wavelet neural network,HWNN). 用小波变换替换式(7)中的傅里叶变换,可得到基于小波变换的超图卷积定义:
$ \begin{split}g\;{\ast }_{\mathcal{G}}\;x=&{{\boldsymbol{\psi}} }_{s}^{\mathcal{G}}(({(}{{{\boldsymbol{\psi}} }_{s}^{\mathcal{G}}})^{-1}g)\odot {(}({{{\boldsymbol{\psi}} }_{s}^{\mathcal{G}}})^{-1}x)) ={{\boldsymbol{\psi}} }_{s}^{\mathcal{G}}{\varLambda }_{\beta }^{\mathcal{G}}{(}{{{\boldsymbol{\psi}} }_{s}^{\mathcal{G}}})^{-1}x, \end{split} $
其中${\boldsymbol{\psi}} _s^\mathcal{G} = {{\boldsymbol{U}}_\mathcal{G}}{\boldsymbol{\varLambda}} _s^\mathcal{G}{\boldsymbol{U}}_\mathcal{G}^{\text{T}}$是缩放参数为$ - s$的小波基的集合,$ {\boldsymbol{\varLambda}}_{s}^{\mathcal{G}}= \text{diag}({{e}}^{-{\lambda }_{\mathcal{G}}^{0}s},{e}^{-{\lambda }_{\mathcal{G}}^{1}s},…,{e}^{-{\lambda }_{\mathcal{G}}^{({N}_{\mathcal{G}}-1)} s}\;\;\;\;) $是热核矩阵且$ {\lambda }_{\mathcal{G}}^{0}\le{\lambda }_{\mathcal{G}}^{1}\le … \le {\lambda }_{\mathcal{G}}^{({N}_{\mathcal{G}}-1)} $是超图拉普拉斯矩阵的特征值;$ {\boldsymbol{\varLambda}}_{\beta }^{\mathcal{G}}=\text {diag}({\beta }_{0}, {\beta }_{1},…,{\beta }_{({N}_{\mathcal{G}}-1)}) $.
根据Stone-Weierstrass定理[120]得到小波变换及其逆变换的近似多项式${({\boldsymbol{\psi}} _s^\mathcal{G})^{ - 1}} \approx (\overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G})' = \displaystyle\sum\limits_{k = 0}^{K'} {{{\overline {\boldsymbol{\varTheta}} '}_k}{{({{\overline {{\varTheta}} }^\mathcal{G}})}^k}} $和${\boldsymbol{\psi}} _s^\mathcal{G} \approx \overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G} = \displaystyle\sum\limits_{k = 0}^K {{{\overline {\boldsymbol{\varTheta}} }_k}{{({{\overline {\boldsymbol{\varTheta}} }^\mathcal{G}})}^k}} $,其中${\mathcal{L}_\mathcal{G}}$是${\overline {\boldsymbol{\varTheta}} ^\mathcal{G}}$的1阶多项式. 这2个近似多项式避免了HGNN中的拉普拉斯分解操作. 将它们代入式(16)后,基于小波变换的超图卷积定义为
$ g\;{ *_\mathcal{G}}\;x = \overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G}{\boldsymbol{\varLambda}} _\beta ^\mathcal{G}(\overline {\boldsymbol{\varTheta}} _{{\text{sum}}}^\mathcal{G})'x. $
最终的HWNN的卷积层定义为
$ {\boldsymbol{Q}}_{\mathcal{G}}^{(l + 1)} = \sigma (\overline {{\boldsymbol{\varTheta}}} _{{\text{sum}}}^{\mathcal{G}} \boldsymbol{\varLambda}_{\boldsymbol{\psi}}^{(l)} \left(\overline {\boldsymbol{\varTheta}}_{\text{sum}}^{{\mathcal{G}}}\right)' {\boldsymbol{Q}}_{\mathcal{G}}^{(l)}{\boldsymbol{W}}^{(l)}_{\psi}), $
其中${\boldsymbol{W}}_\psi ^{(l)}$是特征投影矩阵,${\boldsymbol{\varLambda}} _\psi ^{(l)}$是滤波器.HGNN中的傅里叶变换和HWNN中的小波变换的详细描述如表2所示.
6) HGWNN
HGWNN[22]是另一种引用小波的超图小波神经网络,直接采用截断切比雪夫多项式近似小波系数. 它的卷积层定义为
$ {\boldsymbol{Q}}_{\mathcal{G}}^{(l + 1)} = \sigma \left({{{}^1/{}_2}}{c_{{s_l,0}}}\;{\boldsymbol{I}} + \sum\limits_{k= 1 }^{ K} {c_{{s_l,k}}}{T_k}({\hat{\mathcal{L}}_\mathcal{G}})\right){\boldsymbol{Q}}_\mathcal{G}^{(l)}{{\boldsymbol{\varTheta}}_{{c_s}}} . $
除此之外,HNHN[61]是一种对顶点和超边加入非线性激活函数和归一化操作的超图卷积网络. 超图卷积过程包含归一化节点、更新超边、归一化超边和更新节点4个阶段. 双超图卷积网络(dual hypergraph convolutional network,DualHGCN)[85]通过2个同构超图建模多重二分网络,根据二分网络的特性设计消息内和消息间的传递策略,促进信息共享.Wu等人[106]提出用于属性图学习的双视图超图神经网络(dual-view hypergraph neural network,DHGNNb),采用属性超图和结构超图,分别表示顶点的属性和拓扑结构;再送入共享超边卷积层、特定超边卷积层和注意力层,生成顶点的最终表示.
从上述可知,无向超图卷积神经网络都是以谱理论为基础,通过傅里叶变换或小波变换实现超图卷积,同时也可加入非线性变换、归一化操作、无监督等增强学习能力.
有向超图是一种特殊的超图,有向超边携带重要的方向信息,本节主要介绍以有向超图为拓扑结构的神经网络模型.
1)基于有向超图拉普拉斯的超图神经网络
Tran等人[57]提出基于有向超图拉普拉斯算子的有向超图神经网络,应用于有向超图的半监督学习. 首先用$ {\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{head}}} $和${\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{tail}}}$表示有向超图
$ \vec h_{ij}^{{\text{head}}} = \vec h(v_{{\mathcal{G}_{\mathrm{D}}}}^i,\vec e_j^{{\text{ head}}}) = \left\{ {\begin{aligned}&{1,\quad{\text{if }}v_{{\mathcal{G}_{\mathrm{D}}}}^i \in \vec e_j^{{\text{ head}}},} \\ &{0,\quad{\text{if }}v_{{\mathcal{G}_{\mathrm{D}}}}^i \notin \vec e_j^{{\text{ head}}},} \end{aligned}} \right. $
$ \vec h_{ij}^{{\text{ tail}}} = \vec h(v_{{\mathcal{G}_{\mathrm{D}}}}^i,\vec e_j^{{\text{ tail}}}) = \left\{ {\begin{aligned}&{1,\quad{\text{ if }}v_{{\mathcal{G}_{\mathrm{D}}}}^i \in \vec e_j^{{\text{ tail}}},} \\ &{0,\quad{\text{ if }}v_{{\mathcal{G}_{\mathrm{D}}}}^i \notin \vec e_j^{{\text{ tail}}},} \end{aligned}} \right. $
其中$ \vec h_{ij}^{{\text{head}}} $,$ \vec h_{ij}^{{\text{tail}}} $分别是$ {\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{head}}} $和${\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{tail}}}$的元素,$ i=1,2,\dots, N_{\mathcal{G}_{\mathrm{D}}} $,$ j=1,2,\dots, M_{\mathcal{\mathcal{G}_{\mathrm{D}}}} $.
有向超图拉普拉斯算子$ {\mathcal{L}_{{\mathcal{G}_{\mathrm{D}}}}} $定义为
$ \begin{array}{l}{{\mathcal{L}}}_{{\mathcal{G}}_{D}}={\boldsymbol{I}}-\left(\dfrac{{{\boldsymbol{S}}}^{{{}^1/{}_2}}{\boldsymbol{D}}_{\cal{V}}^{{\text{tail}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{\text{tail}}{\mathcal{W}}_{{\mathcal{G}}_{D}}{\boldsymbol{D}}_{\overrightarrow{e}}^{{{\mathrm{head}}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{{\text{head}}^{\text{T}}}{{\boldsymbol{S}}}^{-{{}^1/{}_2}}}{2}+\right.\\ \left.\dfrac{{{\boldsymbol{S}}}^{-{{}^1/{}_2}}\left({\boldsymbol{D}}_{\cal{V}}^{{\text{tail}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{\text{tail}}{\mathcal{W}}_{{\mathcal{G}}_{D}}{\boldsymbol{D}}_{\overrightarrow{e}}^{{\text{head}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{{\text{head}}^{\text{T}}}\right)^{\text{T}}{{\boldsymbol{S}}}^{{{}^1/{}_2}}}{2}\right). \end{array} $
基于此,有向超图卷积层被定义为
$ \begin{array}{l}{{\boldsymbol{Q}}}_{{\mathcal{G}}_{D}}^{(l+1)}=\sigma \left(\left(\dfrac{{{\boldsymbol{S}}}^{{{}^1/{}_2}}{\boldsymbol{D}}_{\cal{V}}^{{\text{tail}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{\text{tail}}{\mathcal{W}}_{{\mathcal{G}}_{D}}{\boldsymbol{D}}_{\overrightarrow{e}}^{{{\mathrm{head}}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{{\text{head}}^{\text{T}}}{{\boldsymbol{S}}}^{-{{{}^1/{}_2}}}}{2}+\right.\right.\\ \left.\left.\dfrac{{{\boldsymbol{S}}}^{-{{{}^1/{}_2}}}({\boldsymbol{D}}_{\cal{V}}^{{\text{tail}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{\text{tail}}{\mathcal{W}}_{{\mathcal{G}}_{D}}{\boldsymbol{D}}_{\overrightarrow{e}}^{{\text{head}}^{-1}}{{\boldsymbol{H}}}_{{\mathcal{G}}_{D}}^{{\text{head}}^{\text{T}}}{)}^{\text{T}}{{\boldsymbol{S}}}^{{{}^1/{}_2}}}{2}\right){{\boldsymbol{Q}}}_{{\mathcal{G}}_{D}}^{(l)}{{\boldsymbol{\varTheta}} }_{{\mathcal{G}}_{D}}^{(l)}\right),\end{array} $
其中${\boldsymbol{D}}_{\vec e}^{{\text{head}}} $表示超弧的头度矩阵,$ \boldsymbol{D}_{\cal{V}}^{\text{tail}} $表示顶点的尾度矩阵,$S$是对角矩阵,$ {\boldsymbol{\varTheta}} _{{\mathcal{G}_{\mathrm{D}}}}^{(l)} $表示第$l$层可学习的参数矩阵,$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}}^{(l)} $和$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}}^{(l + 1)} $是第$l$和$l + 1$层顶点的隐层表示.
2) DHGCN
在超图卷积运算中融入方向信息,DHGCN的卷积层定义为
$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}}^{(l + 1)} = {\boldsymbol{D}}_{\cal{V}}^{{\text{tai}}{{\text{l}}^{ - 1}}}{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{tail}}}{{\boldsymbol{W}}_{{\mathcal{G}_{\mathrm{D}}}}}{\boldsymbol{D}}_{\vec e}^{{\text{hea}}{{\text{d}}^{ - 1}}}{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{hea}}{{\text{d}}^{\text{T}}}}{\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}}^{(l)}{\boldsymbol{\varTheta}} _{{\mathcal{G}_{\mathrm{D}}}}^{(l)}{\text{. }} $
3)DHConv
有向超图卷积(directed hypergraph convolution,DHConv)[112]包含顶点聚合和有向超边聚合2个阶段. DHConv中有向超图的关联矩阵${\overline {\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}} \in {\mathbb{R}^{{N_{{\mathcal{G}_{\mathrm{D}}}}} \times {M_{{\mathcal{G}_{\mathrm{D}}}}}}}$定义为
$ {\vec{h}}_{ij}=\vec{h}({v}_{{\mathcal{G}}_{D}}^{i},{\vec{e}}_{j})=\left\{\begin{split}& 1,\quad\text{if}\;{v}_{{\mathcal{G}}_{D}}^{i}\in {\mathcal{V}}_{{\mathcal{G}}_{D}}^{\text{head}},\\& \begin{split}-1,\;{\mathrm{if}}\;{v}_{{\mathcal{G}}_{D}}^{i}\in {\mathcal{V}}_{{\mathcal{G}}_{D}}^{\text{tail}},\\ 0,\quad \text{其他},\qquad\;\;\end{split}\end{split}\right. $
其中$ \mathcal{V}_{_{{\mathcal{G}_{\mathrm{D}}}}}^{{\text{head}}} \in {\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}} $表示头顶点集,$ \mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}^{{\text{tail}}} \in {\mathcal{V}_{{\mathcal{G}_{\mathrm{D}}}}} $表示尾顶点集. 用$ {\overline{\boldsymbol{H}}_{{\text{head}}}} = \max ({\overline{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}},0) $和$ {\overline{\boldsymbol{H}}_{{\text{tail}}}} = - \min ({\overline{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}},0) $将$ {{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{D}}}}} $划分为2个关联矩阵,DHConv定义为
$ {{\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}}} = \sum\limits_{k = 0}^{K - 1} {f_{{\text{HT}}}^k({{\boldsymbol{X}}_{{\mathcal{G}_{\mathrm{D}}}}}){{\boldsymbol{\varTheta}}_{{\mathcal{G}_{\mathrm{D}}}}'} } + f_{{\text{TH}}}^k({{\boldsymbol{X}}_{{\mathcal{G}_{\mathrm{D}}}}}){{\boldsymbol{\varTheta}} ''_{{\mathcal{G}_{\mathrm{D}}}}}, $
其中$ {f_{{\text{HT}}}}({\boldsymbol{X}}) = \mathcal{H}_{{\text{head}}}^{\text{T}}(\sigma ({\mathcal{H}_{{\text{head}}}}{\boldsymbol{X}}) \odot ({\mathcal{H}_{{\text{tail}}}}{\boldsymbol{X}})) $是正向聚合,$ {f_{{\text{TH}}}}({\boldsymbol{X}}) = \mathcal{H}_{{\text{tail}}}^{\text{T}}(\sigma ({\mathcal{H}_{{\text{tail}}}}{\boldsymbol{X}}) \odot ({\mathcal{H}_{{\text{head}}}}{\boldsymbol{X}})) $是反向聚合,且$ {\mathcal{H}_{{\text{head}}}} = {\overline{\boldsymbol{H}}_{{{\text{head}}}}{{\boldsymbol{\varTheta}} _{{\text{head}}}}} $,$ {\mathcal{H}_{{\text{tail}}}} = {\overline {\boldsymbol{H}}_{{{\text{tail}}}}{{\boldsymbol{\varTheta}} _{{\text{tail}}}}} $. $ {{\boldsymbol{\varTheta}} '_{{\mathcal{G}_{\mathrm{D}}}}} $,$ {{\boldsymbol{\varTheta}} ''_{{\mathcal{G}_{\mathrm{D}}}}} $,${{\boldsymbol{\varTheta}} _{{\text{tail}}}}$,${{\boldsymbol{\varTheta}} _{{\text{head}}}}$是参数矩阵,$ {{\boldsymbol{X}}_{{\mathcal{G}_{\mathrm{D}}}}} $是有向超图的顶点特征.
4) DHN
DHN通过普通图代表有向超图,每条有向超边${\vec e_i}$表示为图${G_{{\mathcal{E}_{{\mathcal{G}_{\mathrm{D}}}}}}}$中的顶点,超边之间的关系表示为图${G_{{\mathcal{E}_{{\mathcal{G}_{\mathrm{D}}}}}}}$中的边;然后对${G_{{\mathcal{E}_{{\mathcal{G}_{\mathrm{D}}}}}}}$应用现有的图神经网络模型学习有向超边的表示. 每层的DHN定义为
$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{V}}^{(l + 1)} = \sigma ([{\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{V}}^{(l)},{\widetilde {\boldsymbol{H}}_\mathcal{G}}{\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{E}}^{(l)}{\boldsymbol{\varTheta }}_{{\mathcal{G}_{\mathrm{D}}}}^{(l)}]), $
其中$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{E}}^{(l)} $是超边在第$l$层的隐层表示,$ {\widetilde {\boldsymbol{H}}_\mathcal{G}} $是有向超图对应的无向超图的关联矩阵,$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{V}}^{(l)} $和$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{V}}^{(l + 1)} $分别是第$l$和$l + 1$层顶点的隐层表示,$ {\boldsymbol{\varTheta}} _{{\mathcal{G}_{\mathrm{D}}}}^{(l)} $表示可学习的参数矩阵,$[ \cdot , \cdot ]$代表拼接运算. 从式(27)可以看出,当DHN更新顶点的特征表示时,考虑到顶点表示$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{V}}^{(l)} $、有向超边表示$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{D}}}_\mathcal{E}}^{(l)} $和无向超图结构$ {\widetilde {\boldsymbol{H}}_\mathcal{G}} $.
从上述可知,有向超图神经网络模型中的超边都带方向信息. 其中,DHN将有向超图转换为有向图,卷积操作中融入了无向超图的结构. 而其他3种有向超图神经网络都定义了含有方向信息的有向超图关联矩阵,但各自具体的定义方式不同.
在无向超图和有向超图神经网络中,建模数据的超图结构都是静态的,超图结构没有随着顶点特征的更新而动态调整,这会影响模型性能. 为了动态地构建和更新超图结构,学者们设计出动态超图神经网络模型.
1) DHGNNa
最初始构造的超图可能不适合表示最终的数据结构,Jiang等人[52]提出一种动态超图神经网络DHGNNa,包含动态超图构建(dynamic hypergraph construction,DHG)和超图卷积(hypergraph convolution,HGC)2个模块. 其中,HGC模块由顶点卷积VertexConv和超边卷积HyperedgeConv组成. VertexConv聚合超边$e$包含的所有顶点特征,生成超边特征${x_e} = VertexConv({{\boldsymbol{X}}_{{\mathcal{G}_{\text{u}}}}})$;HyperedgeConv引入注意力机制聚合包含顶点$u$的超边特征,生成$u$的新特征${x_u} = HyperedgeConv({x_e})$. 然后根据更新后的顶点特征重新构建超图,堆叠多层DHG和HGC实现动态地构建超图.
2) AdaHGNN
AdaHGNN通过关联矩阵自动学习,构造自适应超图,进而学习高阶语义关系. 在构造超图的过程中,超边被定义为顶点之间的一种抽象关系,可在训练阶段自动地学习.${{\boldsymbol{H}}_{{\mathcal{G}_A}}}$表示自适应超图的关联矩阵,AdaHGNN的超图卷积层定义为
$ {\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{A}}}}^{(l + 1)} = \sigma ({\boldsymbol{D}}_{{\mathcal{V}_{{\mathcal{G}_{\mathrm{A}}}}}}^{ -{{{}^1/{}_2}}}{{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{A}}}}}{\mathcal{W}_{{\mathcal{G}_{\mathrm{A}}}}}{\boldsymbol{D}}_{{\mathcal{E}_{{\mathcal{G}_{\mathrm{A}}}}}}^{ - 1}{\boldsymbol{H}}_{{\mathcal{G}_{\mathrm{A}}}}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{V}_{{\mathcal{G}_{\mathrm{A}}}}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{Q}}_{{\mathcal{G}_{\mathrm{A}}}}^{(l)}{\boldsymbol{\varTheta}} _{{\mathcal{G}_{\mathrm{A}}}}^{(l)}), $
其中$ \boldsymbol{Q}_{\mathcal{G}_{\mathrm{A}}}^{(l)} $和$ \boldsymbol{Q}_{\mathcal{G}_{\mathrm{A}}}^{(l+1)} $是第$l$和$l + 1$层顶点的隐层表示. AdaHGNN与HGNN中构造关联矩阵的方法不同.
3) AHGCN
AHGCN采用自适应超边的构造方法生成基于位置的超图${\mathcal{G}_{{\text{loc}}}}$和基于内容的超图${\mathcal{G}_{{\text{con}}}}$,并引入批量归一化[121]和残差连接加速和稳定训练,AHGCN的超图卷积层定义为
$ {\boldsymbol{Q}}_\mathcal{G}^{(l + 1)} = \sigma (BN({\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{H}}_\mathcal{G}}{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{Q}}_\mathcal{G}^{(l)}{\boldsymbol{\varTheta}} _\mathcal{G}^{(l)}) + {\boldsymbol{Q}}_\mathcal{G}^{(l)}\overline {\boldsymbol{\varTheta}} _\mathcal{G}^{(l)}),{\text{ }} $
其中$BN( \cdot )$是批量归一化, $ {\boldsymbol{Q}}_\mathcal{G}^{(l)} $和$ {\boldsymbol{Q}}_\mathcal{G}^{(l + 1)} $是第$l$和$l + 1$层顶点的隐层表示,${\boldsymbol{ \varTheta}} _\mathcal{G}^{(l)} $和$ \overline {\boldsymbol{\varTheta }}_\mathcal{G}^{(l)} $代表可学习的参数,$ {{\boldsymbol{H}}_\mathcal{G}} $由${\mathcal{G}_{{\text{loc}}}}$和${\mathcal{G}_{{\text{con}}}}$组成.
从上可知,本节提到的超图神经网络通过不同的方式动态地优化超图的建模过程. 其中,DHGNNa利用更新后的顶点特征构建新的超图,AdaHGNN在训练阶段自动地学习顶点间的抽象关系,AHGCN用一种自适应超边的构造方法生成超图. 除此之外,半动态超图神经网络(semi-dynamic hypergraph neural network,SD-HNN)[64]通过构建静态超图和动态超图捕获人体运动学.Zhang等人[97]设计一个可以自适应优化超图结构的超图拉普拉斯适配器 (hypergraph laplacian adaptor,HERALD),动态地更新超图结构. 基于张量的动态超图学习(tensor-based dynamic hypergraph learning,t-DHL)[32]采用张量表示建模超图结构,在学习的过程中动态地更新超图结构.
在介绍本节内容之前,需要明确本节涉及的普通图的含义,具体可参见2.2节中图的定义,每条边仅连接2个顶点. 本节介绍基于普通图的超图神经网络,显示地将图卷积网络扩展到超图上. 它们具有共同的特点,即首先采用特定的方法将超图转换成普通图,然后在这个普通图上执行图卷积神经网络.
1) HyperGCN
HyperGCN是一种对超图运用图卷积网络进行训练的方法. 超图谱理论[122-123]可以将超图转换为一个带权普通图. 具体地,选取每条超边中信号特征最大的2个顶点,形成一条边来表示该超边,超边中的其他顶点被删除,将导致信息的丢失. 因此,HyperGCN引入由删除的顶点组成的中介,将中介中的每个顶点与代表超边选取的2个顶点相连接,使用Chan 等人[124]的方法定义超图拉普拉斯. HyperGCN本质是用带权的普通图代替超图,每条超边退化为带权重的边,然后在带权重的普通图上执行图卷积操作即可.
2) LHCN
线超图卷积网络 (line hypergraph convolution network,LHCN)[56]首次引入超图的线图这一概念,超图的线图是一个普通图. 首先将超图$\mathcal{G}$映射到一个加权带属性的线图$L(\mathcal{G})$,其次对线图$L(\mathcal{G})$执行图卷积操作,超图的线图的卷积定义为
$ {\boldsymbol{Q}}_{L(\mathcal{G})}^{(l + 1)} = \sigma (\widetilde {\boldsymbol{D}}_{L(\mathcal{G})}^{ - {{{}^1/{}_2}}}{\widetilde {\boldsymbol{A}}_{L(\mathcal{G})}}\widetilde {\boldsymbol{D}}_{L(\mathcal{G})}^{ - {{{}^1/{}_2}}}{\boldsymbol{Q}}_{L(\mathcal{G})}^{(l)}{\boldsymbol{\varTheta}} _{L(\mathcal{G})}^{(l)}), $
其中${\widetilde {\boldsymbol{A}}_{L(\mathcal{G})}} = {{\boldsymbol{A}}_{L(\mathcal{G})}} + {{\boldsymbol{I}}}$, $ {\widetilde {\boldsymbol{D}}_{L(\mathcal{G})}} $是${\widetilde {\boldsymbol{A}}_{L(\mathcal{G})}}$的度矩阵, ${\boldsymbol{\varTheta}} _{L(\mathcal{G})}^{(l)}$是第$l$层可学习的参数矩阵, $ {\boldsymbol{Q}}_{L(\mathcal{G})}^{(l)} $和$ {\boldsymbol{Q}}_{L(\mathcal{G})}^{(l + 1)} $分别是第$l$和$l + 1$层的隐层表示. 在$L(\mathcal{G})$上完成图卷积操作后,利用反向映射规则从图卷积结果得到原超图中顶点的标签和属性.
3) HAIN
LHCN需要显式地计算线图,计算成本高,超图注意力同构网络 (hypergraph attention isomorphism network,HAIN)[79]用一种隐式的方式生成线图来解决此问题. 超图$\mathcal{G}$被隐式地生成线图$\widetilde L(\mathcal{G})$,${{\boldsymbol{A}}_{\widetilde L(\mathcal{G})}} = {\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - 1}{{\boldsymbol{H}}_\mathcal{G}}$表示融入超图结构的邻接矩阵. 由于线图中的顶点代表的是超图中的超边,HAIN采用自注意力机制学习线图中顶点的权重,衡量了不同超边对顶点的重要性. 受图同构网络(graph isomorphism network,GIN)[125]的启发,HAIN每层定义为
$ {\boldsymbol{Q}}_{\widetilde L(\mathcal{G})}^{(l + 1)} = \sigma ({{\boldsymbol{H}}_\mathcal{G}}({{\boldsymbol{A}}_{\widetilde L(\mathcal{G})}}{\boldsymbol{Q}}_{{{\widetilde L}_{{\text{att}}}}}^{(l)}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{Q}}_{\widetilde L(\mathcal{G})}^{(l)} + \tau _{\widetilde L(\mathcal{G})}^{(l)}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{Q}}_{\widetilde L(\mathcal{G})}^{(l)}){\boldsymbol{\varTheta}} _{\widetilde L(\mathcal{G})}^{(l)}), $
其中${{\boldsymbol{A}}_{\widetilde L(\mathcal{G})}}{\boldsymbol{Q}}_{{{\widetilde L}_{{\text{att}}}}}^{(l)}$是$\widetilde L(\mathcal{G})$带有自注意力的可学习的邻接矩阵,${\boldsymbol{Q}}_{{{\widetilde L}_{{\text{att}}}}}^{(l)} = {\mathrm{diag}}({\sigma _{{\text{att}}}}({\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{Q}}_{\widetilde L(\mathcal{G})}^{(l)}{\boldsymbol{\theta}} _{{\text{att}}}^{(l)}))$,${\boldsymbol{\theta}} _{att}^{(l)}$是注意力向量.$ {\mathrm{diag}}( \cdot )$代表对角算子,它可将向量转化为对角矩阵, ${\boldsymbol{\varTheta}} _{\widetilde L(\mathcal{G})}^{(l)}$是第$l$层可学习的参数矩阵. $\tau _{\widetilde L(\mathcal{G})}^{(l)}$是一个超参数,表示顶点自身对其邻居聚合表示的重要性.$ {\boldsymbol{Q}}_{\widetilde L(\mathcal{G})}^{(l)} $和$ {\boldsymbol{Q}}_{\widetilde L(\mathcal{G})}^{(l + 1)} $分别是第$l$和$l + 1$层顶点的隐层表示.
4) HI-GCN
超图诱导图卷积网络(hypergraph induced graph convolutional network,HI-GCN)[77]将超图结构植入到图卷积运算中来更新顶点嵌入.HI-GCN的卷积层定义为
$ {\boldsymbol{Q}}_{{G_{\mathrm{c}}}}^{(l + 1)} = \sigma ({\boldsymbol{D}}_{{G_{\mathrm{c}}}}^{ - 1}{{\boldsymbol{H}}_\mathcal{G}}{{\boldsymbol{W}}_{{G_{\mathrm{c}}}}}{\boldsymbol{H}}_\mathcal{G}^{\text{T}}{\boldsymbol{D}}_{{G_{\mathrm{c}}}}^{ - 1}{\boldsymbol{Q}}_{{G_{\mathrm{c}}}}^{(l)}{\boldsymbol{\varTheta}} _{{G_{\mathrm{c}}}}^{(l)}), $
其中${{\boldsymbol{D}}_{{G_{\text{c}}}}}$是${G_{\text{c}}}$的度矩阵,${G_{\text{c}}}$是超图通过团扩张生成的普通图,${{\boldsymbol{W}}_{{G_{\mathrm{c}}}}}$是超边权重矩阵,${\boldsymbol{\varTheta }}_{{G_{\mathrm{c}}}}^{(l)}$是可学习权重矩阵,$ {\boldsymbol{Q}}_{{G_{\mathrm{c}}}}^{(l)} $和$ {\boldsymbol{Q}}_{{G_{\mathrm{c}}}}^{(l + 1)} $是第$l$和$l + 1$层顶点的隐层表示.HI-GCN有效包含了图和超图的信息.
从上述4个模型可知,基于普通图的超图神经网络的共同点是采用一定的策略先将超图转化为普通图,然后以图或者超图作为拓扑结构开发模型. 与上述4个模型类似,神经超链接预测器(neural hyperlink predictor, NHP)[26]采用无向超图和有向超图建模数据,利用团扩张将2种超图转化为普通图后实现超图链路预测.Hou等人[78]提出将加权有向超图转换为加权超边图,而加权超边图就是一个普通图. 除此之外,文献[59]提出一种超图转换为普通图的方法,线扩张 (line expansion,LE),将超图中的顶点-超边对作为普通图中的一个顶点,称之为线顶点,如果线顶点(顶点-超边对)之间有公共的顶点或超边,连接这2个线顶点形成一条边,依次执行,最终得到与之对应的普通图.
注意力机制常用于图神经网络中权衡顶点和边的重要性,取得显著成效. 同样,超图神经网络也可引入注意力机制来体现顶点和超边等的重要性. 本节对现有的超图神经网络以注意力机制为关键技术的模型进行归纳和总结,具体如下所述.
1) Hyper-SAGNN
Hyper-SAGNN是一种用于超图学习的框架,能够处理同构和异构超图、一致和非一致超图,并利用学习到的节点嵌入预测非一致异构超图中的超边.Hyper-SAGNN包含静态嵌入模块、动态嵌入模块和预测模块,在动态嵌入模块中引入多头注意力机制生成动态嵌入.
2) HANs
HANs调整输入的多种模态之间的信息层次,将它们整合到同一信息层次. 采用符号图定义不同的模态之间的公共语义空间,在语义空间中构建共同注意力映射以获取不同模态的联合表示.
3) HyperGAT
HyperGAT设计双注意力机制学习有鉴别力的表示,用于归纳式文本分类和降低计算成本. HyperGAT层定义为
$ \begin{array}{*{20}{c}}{{\boldsymbol{q}}_{{v_i}}^{(l + 1)} = HyperedgeAG{G^{(l + 1)}}({\boldsymbol{q}}_{{v_i}}^{(l)},{\boldsymbol{q}}_{{e_j}}^{(l + 1)}),} \\ {{\boldsymbol{q}}_{{e_i}}^{(l + 1)} = VertexAG{G^{(l + 1)}}({\boldsymbol{q}}_{{v_i}}^{(l)}),} \end{array} $
其中${\boldsymbol{q}}_{{v_i}}^{(l + 1)}$和${\boldsymbol{q}}_{{e_i}}^{(l + 1)}$是第$l + 1$层的顶点${v_i}$和超边${e_i}$的表示. 超边聚合函数$HyperedgeAGG( \cdot )$将包含顶点的所有超边的特征聚合到该顶点,并通过超边级注意力机制为每条超边分配权重以表明每条超边对该顶点的贡献度;节点聚合函数$VertexAGG( \cdot )$将超边中所有顶点的特征聚合到该超边,并通过顶点级注意力机制为每个顶点分配权重以表明每个顶点对于超边的贡献度.
4) HGATs
与HyperGAT类似,超图注意力网络(hypergraph attention networks,HGATs)[21]也是基于双注意力机制的超图注意力网络,包含注意力顶点聚合和注意力超边聚合,在聚合特征的过程中引入注意力机制为顶点和超边分配不同的权重.
5) SAHDL
稀疏注意超图正则化字典学习(sparse attention hypergraph regularized dictionary learning,SAHDL)[73]算法设计出一种新的稀疏注意力机制为每个顶点分配重要性权重,并将权重值融入到构建超图的过程中. 引入稀疏注意力的超图关联矩阵${{\boldsymbol{H}}_{{\mathrm{SA}}}}$定义为
$ \left\{\begin{split} & \mathrm{exp}^{(-dis(v_i,v_u)^2)}\times a_{\mathrm{SA}}^i,\text{ if }v_i\in e_{\mathrm{SA}}, \\ & 0,\qquad\text{ 其他,}\end{split}\right. $
其中${e_{{\mathrm{SA}}}}$代表一条超边,$ {\mathrm{dis}}( \cdot )$表示计算2点之间的距离,${\boldsymbol{a}}_{{\mathrm{SA}}}^i$表示顶点$i$的注意力权重值,可通过公式$ {\mathop {\arg \min }\limits_{{{\boldsymbol{a}}}_{{\mathrm{SA}}}}} f({{\boldsymbol{a}}_{\mathrm{SA}}}) = {\text{ }}||\boldsymbol x - {{\boldsymbol{P}}_K}{{\boldsymbol{a}}_{\mathrm{SA}}}||_\text F^2 + 2\tau ||{{\boldsymbol{a}}_{\mathrm{SA}}}|{|_{{\ell _1}}} $计算得到. $ {{\boldsymbol{a}}_{\mathrm{SA}}} $是权重向量,$ {{\boldsymbol{P}}_K} $表示从K个最近邻居中选取的样本的特征,$ \tau $是控制稀疏性的超参数.
6) STHAN-SR
时空超图注意力网络(spatiotemporal hypergraph attention network,STHAN-SR)[82]利用超图注意力巧妙地将时间霍克斯注意力与空间超图卷积相结合,捕捉时空相关性.${\boldsymbol{H}}_\mathcal{G}^{{\text{att}}}$表示引入注意力机制的超图关联矩阵,STHAN-SR的超图卷积层定义为
$ {\boldsymbol{Q}}_\mathcal{G}^{(l + 1)} = \sigma ({\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{H}}_\mathcal{G}^{{\text{att}}}{\mathcal{W}_\mathcal{G}}{\boldsymbol{D}}_{{\mathcal{E}_\mathcal{G}}}^{ - 1}{({\boldsymbol{H}}_\mathcal{G}^{{\text{att}}})^{\text{T}}}{\boldsymbol{D}}_{{\mathcal{V}_\mathcal{G}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{Q}}_\mathcal{G}^{(l)}{\boldsymbol{\varTheta}} _\mathcal{G}^{(l)}), $
它采用多头注意力机制确保训练稳定.
7) HGTAN
超图三注意力网络(hypergraph tri-attention network,HGTAN)[100]设计了超边内、超边间和超图间3个注意力模块,考虑顶点、超边和超图的重要性. 其中,超边内注意力和超边间注意力与HyperGAT和HGATs类似,但HGTAN中采用超图间注意力模块,权衡不同超图的重要性.
从上述可知,该类超图神经网络模型的共同之处是利用注意力机制突显顶点、超边或超图在消息聚合过程中各自的重要性. 除此之外,有向超图注意力网络(directed hypergraph attention network,DHAT)[112]结合3.1.1.2节中介绍的DHConv和注意力机制挖掘数据间的潜在关系.Sun等人[126]提出一种基于多级超边蒸馏策略的超图神经网络,采用节点级、超边级和语义级注意力学习和更新顶点表示.
本节重点介绍引入RNN构建超图神经网络模型,RNN主要用于捕获时间特性.
1)HGC-RNN
HGC-RNN是一种能够预测结构化时间序列传感器网络数据的模型. 它使用超图卷积提取数据的结构特征,同时采用RNN结构提取数据序列中的时间依赖性. 随时间演化的HGC-RNN层定义为
$ {\boldsymbol{Q}}_\mathcal{G}^{(l + 1)} = {F_{{\text{RNN}}}}({F_{{\text{HGC1}}}}({\boldsymbol{X}}_\mathcal{G}^{(l + 1)}||{\boldsymbol{Q}}_\mathcal{G}^{(l)}),{F_{{\text{HGC2}}}}({\boldsymbol{Q}}_\mathcal{G}^{(l)})), $
其中$ {F_{{\text{RNN}}}} $是循环神经网络,如GRU.$ {F_{{\text{HGC1}}}} $和$ {F_{{\text{HGC2}}}} $是用于更新顶点表示的超图卷积操作.
2) 基于卷积、注意力和GRU的超图神经网络
Xia等人[127]开发了一种基于时间的超图神经网络,集成了超图卷积、GRU和注意力机制,有效地学习超图中异构数据之间的高阶相关性.
上述2种超图神经网络有效地组合RNN与超图卷积. 与它们不同的是3.1.3节中的HGTAN结合RNN和超图注意力机制,使用GRU从时间序列数据中捕捉顶点的时间动态性,再将其送入HGTAN更新节点的表示. 1)和2)都充分地考虑了数据间的时间特性和拓扑结构.
受生成对抗网络的启发,研究者引入生成对抗网络设计出超图神经网络模型,本节详细地介绍这类网络.
1) HGGAN
HGGAN提出用交互式超边神经元模块(interactive hyperedge neurons module,IHEN)作为生成器捕获数据间的复杂关系;鉴别器是MLP.IHEN的定义为:
$ \begin{array}{c}{{\boldsymbol{X}}}_{\mathcal{V}}^{(l+1)}=\sigma ({{\boldsymbol{H}}}_{\mathcal{G}}{{\boldsymbol{X}}}_{{\mathcal{E}}}^{(l)}{\mathcal{W}}_{{\mathcal{E}}}^{(l)}+{\lambda }_{\text{IHEN}}{{\boldsymbol{X}}}_{\mathcal{V}}^{(l)}{\mathcal{W}}_{\mathcal{V}}^{(l)}),\\ {{\boldsymbol{X}}}_{{\mathcal{E}}}^{(l+1)}=\sigma ({{\boldsymbol{H}}}_{\mathcal{G}}^{\text{T}}{{\boldsymbol{X}}}_{\mathcal{V}}^{(l)}{\mathcal{W}}_{\mathcal{V}}^{(l)}+{\lambda }_{\text{IHEN}}{{\boldsymbol{X}}}_{{\mathcal{E}}}^{(l)}{\mathcal{W}}_{{\mathcal{E}}}^{(l)}),\end{array} $
其中$ {\boldsymbol{X}}_\mathcal{V}^{(l + 1)} $和$ {\boldsymbol{X}}_\mathcal{E}^{(l + 1)} $分别是第$l + 1$层的节点和超边的特征,$ \mathcal{W}_\mathcal{V}^{(l)} $和$ \mathcal{W}_\mathcal{E}^{(l)} $分别是第$l$层的节点和超边的权重矩阵,$ {\lambda _{{\text{IHEN}}}} $是超参数.
2) MRL-AHF
MRL-AHF利用多模态间的互补和多模态内的交互,提升表示学习能力和多模态融合性能. 潜在表示分别送入编码器EA和EB得到表示${\boldsymbol{X}}_\mathcal{V}^{{\text{EA}}}$和${\boldsymbol{X}}_\mathcal{V}^{{\text{EB}}}$,且生成的2个超图用关联矩阵${{\boldsymbol{H}}_{{\text{EA}}}}$和${{\boldsymbol{H}}_{{\text{EB}}}}$表示. 通过对抗训练策略实现超图融合,得到顶点的特征.
$ {{\boldsymbol{Q}}_{{\text{fuse}}}} = {\boldsymbol{D}}_{{\mathcal{V}_{{\text{EA}}}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{H}}_{{\text{EA}}}}{\boldsymbol{D}}_{{\mathcal{V}_{{\text{EA}}}}}^{-{{}^1/{}_2}}{\overline {\boldsymbol{X}}_{{\text{EA}}}} + {\boldsymbol{D}}_{{\mathcal{V}_{{\text{EB}}}}}^{ - {{{}^1/{}_2}}}{{\boldsymbol{H}}_{{\text{EB}}}}{\boldsymbol{D}}_{{\mathcal{V}_{{\text{EB}}}}}^{ - {{{{}^1/{}_2}}}}{\overline {\boldsymbol{X}}_{{\text{EB}}}}, $
其中$ {{\boldsymbol{D}}_{{\mathcal{V}_{{\text{EA}}}}}} $和$ {{\boldsymbol{D}}_{{\mathcal{V}_{{\text{EB}}}}}} $分别是${{\boldsymbol{H}}_{{\text{EA}}}}$和$ {{\boldsymbol{H}}_{{\text{EB}}}} $的顶点度矩阵,$ {\overline {\boldsymbol{X}}_{{\text{EA}}}} = {\boldsymbol{D}}_{{\mathcal{E}_{{\text{EA}}}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{H}}_{{\text{EA}}}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{E}_{{\text{EA}}}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{X}}_\mathcal{V}^{{\text{EA}}} $,$ {\overline {\boldsymbol{X}}_{{\text{EB}}}} = {\boldsymbol{D}}_{{\mathcal{E}_{{\text{EB}}}}}^{ - {{{}^1/{}_2}}}{\boldsymbol{H}}_{{\text{EB}}}^{\text{T}}{\boldsymbol{D}}_{{\mathcal{E}_{{\text{EB}}}}}^{-{{{}^1/{}_2}}}{\boldsymbol{X}}_\mathcal{V}^{{\text{EB}}}{{\boldsymbol{\varTheta}} _{{\text{EB}}}} $.$ {{\boldsymbol{D}}_{{\mathcal{E}_{{\text{EA}}}}}} $和$ {{\boldsymbol{D}}_{{\mathcal{E}_{{\text{EB}}}}}} $分别是${{\boldsymbol{H}}_{{\text{EA}}}}$和${{\boldsymbol{H}}_{{\text{EB}}}}$的超边度矩阵,$ {{\boldsymbol{\varTheta}} _{EB}} $是权重矩阵.
从上述可知,HGGAN和MRL-AHF巧妙地在生成对抗网络中融入超图结构,提升了模型学习表示的能力.
文献[128]通过低秩滤波器设计高效的超图卷积. 非线性超图扩散 (nonlinear hypergraph diffusion, HyperND)[87] 沿着超图结构传播特征和标签. 受双曲图卷积神经网络[129]的启发,Li等人[103]提出一种带有预训练阶段的双曲超图表示学习方法(hyperbolic hypergraph representation learning method for sequential recommenddation,H2SeqRec)有效地缓解超图的稀疏性和长尾数据分布问题,通过双曲空间超图神经网络学习动态嵌入. DHT(dual hypergraph transformation)[96]是一种边表示学习框架. Srinivasan等人[130]设计了一种用于学习顶点、超边和完全超图表示的超图神经网络. 受GCNII[131]的启发,残差增强多超图神经网络(residual enhanced multihypergraph neural network,ResMultiHGNN)[93]缓解加深HGNN引起的过平滑等问题. Lin等人[113]使用采样超边等方式设计出一种深层超图神经网络(deep gypergraph neural network,DeepHGNN). 深度超边 (deep hyperedges,DHE)[132]利用上下文和集合成员对超图进行转导和归纳式学习. G-MPNN[70]统一不同结构的消息传递神经网络(message passing neural network,MPNN)[133],对多关系有序超图有强大的归纳能力. UniGNN[95]是一种将经典的GNNs,如GCN,GAT(graph attention network)等推广到超图上的统一框架. HyperSAGE[74]受GraphSAGE(graph sample and aggregate) [134]的启发,设计出一种归纳式超图学习框架. HyperGene[92]是一种基于GNNs的超图预训练框架. NetVec[86]是一个能够提高运行速度,用于解决大型超图的多层超图嵌入框架. AllSet [99]将传播规则描述为2个多集函数的组合,并将多集函数的学习问题与超图神经网络联系起来.
现实生活中很多数据间的交互关系可以很自然地用超图表示,虽然超图神经网络的发展历程较短,但它已经被广泛地应用到不同领域的任务中并取得成功. 本节详细地介绍超图神经网络在各领域中的实际应用.
在引文网络中,超图可用于建模论文的引用关系和论文的合著关系. 引文网络分类任务属于节点级任务,常用的数据集有Cora,DBLP,Pubmed,Citeseer,具体描述如表3所示. HGNN, HyperGCN, DHGNNa, HyperSAGE, LHCN, iHGAN[80],HNHN ,HyperND等都可用于引文网络分类. 表4列举了应用于引文网络分类的超图神经网络模型及其采用的数据集和评价指标.
对于半监督节点分类任务,表5罗列了3种超图神经网络模型与经典的GCN在Cora上的比较,实验中的训练集和测试集数目分别为140和1000.其中DHGNNa和iHGAN的准确度分别比GCN提升了1%和1.4%,展现了超图神经网络的有效性.
多超图神经网络(multi-hypergraph neural network,MHGNNa)[51]利用生理信号识别人类情绪,每种类型的生理信号建模为一个超图,超图中的顶点是一个受试对象和刺激物的二元组,超边代表受试者之间的相关性. 若存在多种生理信号,则采用多个超图建模. 每个超图上执行卷积后融合多个超图卷积的输出结果,得到顶点的最终表示. 文献[53]中提出一种利用生理信号识别情绪的边加权超图神经网络,在训练的过程中同时学习超边权重和滤波器参数.
序列超图增强下一项推荐(sequential hypergraphs to enhance next-item recommendation,HyperRec)[23]是一种端到端的框架,该框架利用序列超图描述项目之间的短期相关性,结合超图卷积层生成有效的项目动态嵌入. H2SeqRec在双曲空间中设计超图卷积,从年度、季度和月度多个时间维度学习项目的嵌入. DHCF是双通道协同过滤框架,它采用分治策略学习用户和项目的不同表示. 用户间和项目间的高阶相关性建模为用户超图和项目超图,并利用超图卷积学习用户和项目的表示. 双通道的设置使得在集成用户和项目一起推荐的同时,又保持各自的属性.SHCN 是基于用户评论的信号超图卷积网络,设计了符号超边卷积操作,通过自定义的邻域聚合策略有效地传递用户偏好,改进特征感知的推荐性能.
DHCN, SHARE [90], HGNNA[110]是基于会话推荐的超图神经网络模型. DHCN是结合超图和超图线图设计的双通道超图卷积网络,其中超图卷积负责捕获项目级的高阶关系,超图的线图卷积负责学习会话级关系. S2-DHCN是加入了自监督学习的DHCN变体. SHARE用一个超图表示每个会话,项目是超图中的顶点,超边连接一个上下文窗口中所有的项目. 引入超图注意力机制来衡量项目对会话的重要性和会话对项目的影响. HGNNA是基于超图神经网络和注意力机制的推荐模型. HGNNA首先利用超图神经网络学习项目间的关联,然后引入自注意力机制聚合会话信息,最后通过图注意机制挖掘出会话间的相关性.
STAMP(short-term attention/memory priority)[135]和SR-GNN(session-based recommendation with graph neural networks)[136]是2个先进的基线模型,分别通过引入注意力机制和GNNs来增强基于会话的推荐.表6记录了基于会话推荐常用的数据集YooChoose和Diginetica. 从表7中可以看出,超图神经网络模型SHARE,DHCN, S2-DHCN优于SR-GNN,在Yoochoose1/4上平均倒数排名(mean reciprocal ranking,MRR)的MRR@K值分别提升了0.22%,4.83 %,8.24%;在Diginetica上MRR@K值分别提升了0.46%,12.99%,13.35%. 而HGNNA的性能也优于STAMP,证明了超图神经网络模型在该任务中的成效显著.
MHCN,HASRE[109],SHGCN[111]都是用于社会推荐的超图神经网络模型.MHCN是多通道超图卷积网络,它利用多通道,如社交通道、联合通道和购买通道分别编码用户的高阶关系,在超图卷积网络的训练中引入自监督学习增强推荐性能. HASRE是基于超图注意力网络的社会推荐算法,超图建模用户之间的高阶关系,引入图注意力机制为用户的好友分配合适的权重,反映出好友对用户的不同影响. SHGCN主要解决非同质社会推荐问题,利用超图表示复杂的三元社交关系(用户-用户-项目),设计超图卷积网络学习社会影响和用户偏好.
LightGCN[137]是一个非常具有竞争力的基线模型,它对GCN的内部结构进行了调整,大多数情况下在社会推荐任务中表现出最好或次优的性能. LastFM,Douban,Yelp是常用于社会推荐的数据集,详情如表8所示. 表9记录了LightGCN与超图神经网络模型DHCF,MHCN,HASRE的比较,实验结果引用自HASRE,故这里的训练集、验证集和测试集的占比与HASRE一样,分别为80%,10%,10%. 除了DHCF,其他3个超图神经网络模型的性能都优于LightGCN. 其中,HASRE在3个数据集中均表现出最佳的效果,展现了超图神经网络模型在社会推荐任务中具有显著的效果.
群体推荐的分层超图学习框架HHGR[107]分别在用户级和组级超图上设计超图卷积网络,捕获用户之间的组内和组间的交互. HyperTeNet[108]用于实现任务中的个性化列表延续任务,构建3一致超图,采用Hyper-SAGNN学习用户、项目和列表之间的三元关系. Liu等人[89]设计的知识感知超图神经网络(knowledge-aware hypergraph neural network,KHNN),从知识图谱学习用户、项目和实体之间的高阶关系. 表10介绍了应用于推荐的超图神经网络模型及其采用的数据集和评价指标.
Wadhwa等人[138]提出一种基于超图卷积的图像补全技术,从输入的数据中学习超图结构. 超图卷积中代表超图结构的关联矩阵包含每个顶点对超边的贡献信息,由空间特征的互相关计算贡献权重. HGCNN(hypergraph convolutional neural networks)[63]是解决3D人脸反欺骗问题的超图卷积神经网络. 每张人脸构建一个超图,利用超图卷积学习深度特征和纹理特征,将它们拼接起来作为用于分类的最终表示. 谱-空间超图卷积神经网络(spectral-spatial hypergraph convolutional neural network,S2HCN)[105]是用于高光谱图像分类的超图卷积神经网络,首先从高光谱图像中挖掘谱特征和空间特征,然后分别基于这2个特征构建超图,并将这些超图融合为一个超图,最后将超图和初始高光谱图像作为超图卷积的输入. 多粒度超图(multi-granular hypergraph,MGH)学习框架[98]能够同时运用时间和空间线索完成视频身份识别,故超图神经网络可用于探索视频序列中的多粒度时空线索之间的高阶相关性.
HI-GCN和AdaHGNN用于多标签图像分类任务.HI-GCN采用数据驱动的方式为标签构建一个自适应超图,再利用超图诱导图卷积网络更新标签嵌入. AdaHGNN用自动学习关联矩阵的方法来构造自适应超图,利用超图神经网络学习多标签之间的高阶语义关系. HI-GCN和AdaHGNN构建的超图中,每个标签作为一个顶点,但定义超边的方法不同. Dang等人[139]设计一种用于解决带噪声的标签学习问题的超图神经网络.
Zheng等人[140]为视觉事件推理任务设计了一种多模态模型,称为超图增强图推理. 该模型通过有效地集成图卷积、超图卷积和自注意力机制,提升模型的视觉推理和视觉叙述能力. 超图构建采用KNN算法,其中顶点代表视觉特征和文本特征,超图卷积操作与HGNN类似.
多模式学习的主要任务是处理来自多个不同来源的信息,视觉问答(VQA)是一种典型的多模式学习任务,它的任务目标是回答关于图像场景的文本问题,需要处理包括图像和文本的信息.HANs是用于多模态学习的超图注意力网络.
Hyper-GNN[84]是基于骨骼的动作识别的超图神经网络,通过三流超图神经网络动态地融合关节流、骨骼流和运动流3种特征,并充分地利用它们之间的互补性和多样性. SD-HNN通过半动态超图神经网络从2D姿势重新估计人体的3D姿势,用超图建模人体,使用静态超图和动态超图分别表示关节之间的局部和全局运动约束,有效地刻画了人体关节之间的复杂运动学关系.
视觉对象分类任务中,每个对象作为顶点,采用KNN算法生成超边,运用HGNN和HGATs实现对象识别任务. HGNN完全依赖超图结构,没有突显出顶点和超边的重要性;HGATs不依赖于超图结构,对顶点和超边赋予不同的重要性,可以用于归纳学习. 对于多模态数据,MultiHGNN拼接(采用均值)的是卷积结果,而HGNN拼接的是超图结构. HGWNN是引入小波变换设计的超图小波神经网络,采用最远点采样(FPS)算法和球查询方法构建超图. 多尺度超图神经网络(multi-scale hypergraph neural network,MHGNNb)[91]的超图卷积操作与HGNN类似,但它融合了每层超图卷积的顶点特征;而HGNN仅使用最后一层特征,未考虑每一层的顶点特征信息,显得信息比较单一. DeepHGNN是一种加深HGNN的超图神经网络.
NTU和ModelNet40是超图神经网络常用于视觉对象分类的数据集,具体细节如表11所示. 表12记录了6个超图神经网络与GCN在2个数据集上的实验结果,可以观察到超图神经网络模型在NTU上的分类准确率都高于84%,比GCN至少提高了8%, 其中HGWNN和ResMultiHGNN分类准确率分别提高了15%和16%. 同样地,超图神经网络模型在ModelNet40上的分类准确率也都优于GCN,充分地展现出超图神经网络在这类任务中的优势. 表13总结了应用于计算机视觉的超图神经网络模型及其采用的数据集和评价指标.
HyperGAT是针对文本分类任务设计的基于双注意力机制的超图注意力网络,使用文档级超图建模每个文本文档,顶点表示文档中的单词,存在顺序超边和语义超边. 采用顶点级和边级注意力学习高阶关系和有判别力的表示. 该模型适用于归纳式训练.
生物医学事件抽取是涉及生物医学概念的一项抽取任务,需要从生物医学的上下文本中提取有用的基本信息. Zhao等人[141]提出一个用于解决该任务的端到端的文档2级联合框架,通过堆叠 HANN 层,有效地建模生物医学文档中局部和全局上下文之间的相互关联.
DHGCN是在超图卷积操作中融入方向信息的有向超图卷积网络,它很自然地将超图卷积扩展到有向超图,提出基于有向超图卷积网络的多跳知识库问答模型2HR-DR. 受CompGCN [142]和递归超图的启发,Yadati等人[143]引入递归超图建模知识库,提出递归超图网络(recursive hypergraph network, RecHyperNet),用于知识库问答任务. Han等人[144]提出基于文本增强知识图谱的开放领域问答模型,利用文本语义信息增强实体表示和文本结构补充知识库中的关系. 其中每个文本视为一条超边,用于连接该文本中的实体,然后对文本超边形成的超图执行超图卷积,更新实体信息. Fatemi等人[65]提出HSimplE和HypE直接对知识超图进行链路预测.
知识库问答是自然语言处理中的一种任务,常用的传统模型有KV-MemNet(key-value memory network)[145]和 GraftNet(graphs of relations among facts and text networks)[146],前者将知识库事实以键-值的结构存储;后者用文本信息和实体一起构建图,并将 GCN 应用于推理. 常用数据集MetaQA和WebQuestionsSP的描述如表14所示. 虽然传统模型有各自的优势,但从表15可以看出超图神经网络模型2HR-DR 和RecHyperNet的性能更优. 特别地,RecHyperNet在2个数据集上比GraftNet分别提升了2.1个百分点和1.6个百分点,表明超图神经网络模型的可行性和有效性. 实验结果来自RecHyperNet中的实验,故具有可比性. 表16总结了应用于自然语言处理的超图神经网络模型及其采用的数据集和评价指标.
时空超图卷积网络STHGCN、时空超图注意力网络STHAN-SR和超图三注意力网络HGTAN都是用于预测股票趋势的模型. 具体地,STHGCN用超图表示股票间的产业关系,设计时空超图卷积学习股票价格的时间演化和股票之间的关系. STHAN- SR将股票预测任务刻画为一个学习排序问题,结合超图卷积和时间霍克斯注意机制探索股票之间的空间复杂关系和时间依赖性. HGTAN引入分层注意力机制,设计超边内、超边间和超图间的注意力分别衡量顶点、超边和超图在股票间信息传递中的重要性,从而挖掘出股票运动的潜在关系. 表17记录了应用于预测股票趋势的超图神经网络模型及其采用的数据集和评价指标.
Hyper-SAGNN能够学习有效的细胞嵌入,有助于理解高阶基因组的细胞间变化. NHP采用有向超图和无向超图建模化学反应网络,2个变体NHP-D(有向)和NHP-U(无向)用于化学反应预测,通过团扩张将超图转化为普通图后实现超图链路预测. Hou等人[78]用加权有向超图建模细胞之间的多对多关系,将加权有向超图转换为超边图,并在超边图上运用 MCL (Markov cluster algorithm)[147]和Louvain[148]社区检测算法.
HGGAN是用于分析阿尔茨海默病的超图生成对抗网络. MRL-AHF是用于诊断阿尔茨海默病的框架,由多模态表示学习和对抗性超图融合组成. HUNet(hypergraph U-Net)[68]利用超图结构学习数据样本的嵌入和高阶关系,应用于脑图的嵌入和分类. Banka等人[69]设计超图自编码器学习大脑区域间的高阶关系,实现脑状态分类. Madine等人[149]设计了诊断自闭症谱系障碍的超图神经网络,提出用无监督多核学习增强超图神经网络的学习能力. 表18描述了应用于生化医疗领域的超图神经网络模型及其采用的数据集和评价指标.
地理-语义-时间超图卷积网络GST-HCN用于预测交通流,利用地理-时间图卷积和语义-时间超图卷积联合学习地理、语义和时间的高阶相关性. 动态时空超图图卷积网络(dynamic spatio-temporal hypergraph graph convolution network,DSTHGCN)[88]用于预测地铁客流,超图建模轨道交通数据,充分地考虑了地铁的拓扑结构和乘客出行方式的多样化,设计超图卷积和时间卷积学习顶点的时空特征,进而提高预测性能. 鉴于有向超图可以同时刻画空间信息和有向关系, DHAT采用了有向超图建模道路网络,基于有向超图设计有向超图卷积挖掘交通序列之间的空间关系,并结合注意力机制实现有效的交通预测. 表19描述了应用于交通预测的超图神经网络模型及其采用的数据集和评价指标.
化学机械抛光(chemical mechanical planarization,CMP)是半导体行业中的一种关键工艺,其表面平整度在很大程度上影响制造质量. 材料去除率(material removal rate, MRR)在这个过程处于重要的地位. Xia等人[127]开发了一种基于时间的超图卷积网络,用于预测CMP过程中的MRR,通过超图卷积、超图注意力和GRU学习异构数据之间的相关性.
近3年,在图神经网络的成功和超图建模的优势的推动下,超图神经网络快速发展并取得显著的成果,但由于超图神经网络的研究还处于新起阶段,因此存在需要迫切解决的问题,并可将其作为进一步研究的方向.
采用图建模数据在设计图神经网络时扮演着重要的角色,与构建图神经网络一样,构建超图在设计超图神经网络模型的过程中也具有重要的地位,会直接影响模型性能. 现有的超图神经网络利用超图建模数据间的高阶复杂关系,构建超图的常用方法有基于距离、基于表示、基于属性和基于网络的方法[32]. 然而,这些方法所构建出的超图结构是静态的,在学习的过程中没有随着顶点表示的更新而动态调整. 虽然DHGNNa和AHGCN等模型中已经考虑动态地构造超图,但是都有各自的局限性. 如在加入新数据时需要重新训练等问题. 现有的适用于超图神经网络的超图构造方法的研究工作比较少,故对于如何构造切实有助于从结构上提升模型性能的超图十分必要.
图神经网络关于深层模型的探索已取得令人鼓舞的成果,如JK-Nets(jumping knowledge networks)[150], DropEdge[151],GCNII等. 尽管近3年来超图神经网络在各个领域都有广泛的应用,但它们大多是浅层模型. 在不牺牲性能的前提下,关于深层模型的研究很少,如ResMultiHGNN,DeepHGNN,UniGNN,其中ResMultiHGNN和UniGCN引入GCNII中的方法,将其直接地从图结构推广到超图上. DeepHGNN提出用采样超边等方式加深网络,但局限于加深的模型有限,并未提出一个超图神经网络通用的深层框架. 因此,在加深的过程中是否会遇到与加深GNNs时相同的问题,如过拟合、过平滑等;是否会出现GNNs中未遇到的其他问题;通过什么技术可以有效地缓解出现的问题都需要在将来加以解决和研究.
现有的超图神经网络模型大多数是显式或隐式地推广图神经网络模型到超图建模的数据上. 图和超图刻画和表示数据相关性不同,但现有探讨超图神经网络模型的文献中涉及模型理论研究的很少. 这些文献的重点是在具体任务中提升模型性能,少有从理论的角度解释模型. 超图除了拥有灵活的建模能力外,还具有其他属性和性质,因此,如何使用这些属性和性质构建更好的超图神经网络,如何将超图理论作为工具探究和解释模型也是值得研究的工作.
现实生活中存在很多关系复杂的网络,虽然现有超图神经网络的应用涉及社交网络、推荐系统、计算机视觉、自然语言处理、金融、生物化学和交通预测等,但在具体的领域中涉及的任务较少且单一,如金融领域仅考虑了股票趋势预测的任务. 因此,研究如何将超图神经网络深度地应用于各个领域也十分有意义.
从2019年Feng等人[19]首次提出的超图神经网络HGNN,到DHGCN,DHGNNa,HWNN,HyperGAT, HGC-RNN,HGGAN, MHCN等不同模型. 这些模型不断丰富和完善超图神经网络家族. 虽然超图神经网络发展的时间短、起步晚,但它借助可以挖掘数据间的潜在高阶关系吸引越来越多的研究者从事这方面的探索和钻研. 他们主要的工作是优化和改进模型算法、超图结构,推动超图神经网络应用于各个领域并取得不俗的表现. 因此,通过阅读大量文献,本文全面地回顾和综述了超图神经网络模型的研究历程.
本文概括和分析了近3年出现的超图神经网络模型和框架,将超图神经网络分为超图卷积神经网络、超图循环神经网络和超图生成对抗网络3类. 在第3节中详细地介绍了这3类中代表性的模型算法并做比较. 第4节中归纳了超图神经网络在社交网络、推荐系统、计算机视觉和自然语言处理等领域的应用,并列举了所采用的数据集和评价指标. 通过第5节探讨了存在于超图神经网络中一些需要解决的关键研究问题.
总之,超图神经网络作为一种新兴的研究方向,由于其发展历程较短,现有的模型存在自身的不足和局限性,有待改进和提升的空间很大. 因此,其具有丰富的可研究内容. 未来,随着更多研究者的加入,这些问题会得到不断改善和解决,同时会开发出更多优秀的模型来扩充超图神经网络体系,拓展其应用的广度和深度.
作者贡献声明:林晶晶收集和分析相关参考文献,负责论文撰写和校对;冶忠林、赵海兴提出指导意见并修改论文;李卓然收集和整理相关文献.

参考文献 引证文献
排序方式:
1
Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2012, 60(6): 84−90
2
Elman J L. Distributed representations, simple recurrent networks, and grammatical structure[J]. Machine Learning, 1991, 7(2): 195−225.
3
LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436−444
4
Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735−1780
5
Chung J, Gülçehre Ç, Cho K, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[J]. arXiv preprint, arXiv: 1412. 3555, 2014
6
徐冰冰,岑科廷,黄俊杰,等. 图卷积神经网络综述[J]. 计算机学报,2020,43(5):755−780
Xu Bingbing, Cen Keting, Huang Junjie, et al. A survey on graph convolutional neural network[J]. Chinese Journal of Computers, 2020, 43(5): 755−780(in Chinese)
7
Bruna J, Zaremba W, Szlam A, et al. Spectral networks and locally connected networks on graphs[J]. arXiv preprint, arXiv: 1312. 6203, 2014
8
马帅,刘建伟,左信. 图神经网络综述[J]. 计算机研究与发展,2022,59(1):47−80
Ma Shuai, Liu Jianwei, Zuo Xin. Survey on graph neural network[J]. Journal of Computer Research and Development, 2022, 59(1): 47−80(in Chinese)
9
李涵,严明玉,吕征阳,等. 图神经网络加速结构综述[J]. 计算机研究与发展,2021,58(6):1204−1229
Li Han, Yan Mingyu, Lü Zhengyang, et al. Survey on graph neural network acceleration architectures[J]. Journal of Computer Research and Development, 2021, 58(6): 1204−1229(in Chinese)
10
Velickovic P, Cucurull G, Casanova A, et al. Graph attention networks[J]. arXiv preprint, arXiv: 1710. 10903, 2017
11
Ye Zhonglin, Zhao Haixing, Zhu Yu, et al. HSNR: A network representation learning algorithm using hierarchical structure embedding[J]. Chinese Journal of Electronics, 2019, 29(6): 1141−1152
12
Kipf T N, Welling M. Variational graph auto-encoders[J]. arXiv preprint, arXiv: 1611. 07308, 2016
13
Zhang Muhan, Chen Yixin. Link prediction based on graph neural networks [C] //Proc of the 32nd Int Conf on Neural Information Processing Systems. Red Hook: Curran Associates Inc, 2018: 5171–5181
14
Ye Zhonglin, Zhao Haixing, Zhang Kecheng, et al. Tri-party deep network representation learning using inductive matrix completion[J]. Journal of Central South University, 2019, 26(10): 2746−2758
15
Hamaguchi T, Oiwa H, Shimbo M, et al. Knowledge transfer for out-of-knowledge-base entities: A graph neural network approach[J]. arXiv preprint, arXiv: 1706. 05674, 2017
16
Marcheggiani D, Bastings J, Titov I. Exploiting semantics in neural machine translation with graph convolutional networks[C] //Proc of the 16th Annual Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg, PA: ACL, 2018: 486−492
17
Wu Zonghan, Pan Shirui, Chen Fengwen, et al. A comprehensive survey on graph neural networks[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 32(1): 4−24
18
Han Yi, Zhou Bin, Pei Jian, et al. Understanding importance of collaborations in co-authorship networks: A supportiveness analysis approach [C] //Proc of the 9th SIAM Int Conf on Data Mining. Philadelphia, PA: SIAM, 2009: 1112−1123
19
Feng Yifan, You Haoxuan, Zhang Zizhao, et al. Hypergraph neural networks [C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 3558−3565
20
Yadati N, Nimishakavi M, Yadav P, et al. HyperGCN: A new method of training graph convolutional networks on hypergraphs [C] // Proc of the 33rd Neural Information Processing Systems. New York: Curran Associates Inc, 2019: 1511−1522
21
Chen Chaofan, Cheng Zelei, Li Zuotian, et al. Hypergraph attention networks[C] //Proc of the 19th Int Conf on Trust, Security and Privacy in Computing and Communications (TrustCom). Piscataway, NJ: IEEE, 2020: 1560−1565
22
Nong Liping, Wang Junyi, Lin Jiming, et al. Hypergraph wavelet neural networks for 3D object classification[J]. Neurocomputing, 2021, 463: 580−595
23
Wang Jianling, Ding Kaize, Hong Liangjie, et al. Next-item recommendation with sequential hypergraphs [C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 1101−1110
24
Chen Xu, Xiong Kun, Zhang Yongfeng, et al. Neural feature-aware recommendation with signed hypergraph convolutional network[J]. ACM Transactions on Information Systems, 2020, 39(1): 1−22
25
Zhang Ruochi, Zou Yuesong, Ma Jian. Hyper-SAGNN: A self-attention based graph neural network for hypergraphs[J]. arXiv preprint, arXiv: 1911. 02613, 2019
26
Yadati N, Nitin V, Nimishakavi M, et al. NHP: Neural hypergraph link prediction [C] //Proc of the 29th ACM Int Conf on Information & Knowledge Management. New York: ACM, 2020: 1705−1714
27
Zhou Jie, Cui Ganqu, Hu Shengding, et al. Graph neural networks: A review of methods and applications[J]. AI Open, 2020, 1: 57−81
28
周飞燕,金林鹏,董军. 卷积神经网络研究综述[J]. 计算机学报,2017,40(6):1229−1251
Zhou Feiyan, Jin Linpeng, Dong Jun. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017, 40(6): 1229−1251(in Chinese)
29
Asif N A, Sarker Y, Chakrabortty P K, et al. Graph neural network: A comprehensive review on non-Euclidean space[J]. IEEE Access, 2021, 9: 60588−60606
30
Malekzadeh M, Hajibabaee P, Heidari M, et al. Review of graph neural network in text classification [C] //Proc of the 12th Annual Ubiquitous Computing, Electronics & Mobile Communication Conf. Piscataway, NJ: IEEE, 2021: 0084−0091
31
Wu Shiwen, Zhang Wentao, Sun Fei, et al. Graph neural networks in recommender systems: A survey[J]. arXiv preprint, arXiv: 2011. 02260, 2020
32
Gao Yue, Zhang Zizhao, Lin Haojie, et al. Hypergraph learning: Methods and practices[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 44(5): 2548−2566
33
Li Yikang, Ouyang Wanli, Zhou Bolei, et al. Factorizable net: An efficient subgraph-based framework for scene graph generation[J]. arXiv preprint, arXiv: 1806. 11538, 2018
34
Yan Sijie, Xiong Yuanjun, Lin Dahua. Spatial temporal graph convolutional networks for skeleton-based action recognition[C] //Proc of the 32nd AAAI Conf on Artificial Intelligence. Palo Alto, CA : AAAI, 2018: 7444–7452
35
Wang Yue, Sun Yongbin, Liu Ziwei, et al. Dynamic graph cnn for learning on point clouds[J]. ACM Transactions on Graphics, 2019, 38(5): 1−12
36
Bastings J, Titov I, Aziz W, et al. Graph convolutional encoders for syntax-aware neural machine translation [C]// Proc of the 2017 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA : ACL, 2017: 1957−1967
37
Beck D, Haffari G, Cohn T. Graph-to-sequence learning using gated graph neural networks[C] //Proc of the 56th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2018: 273−283
38
Yu Ting, Yin Haoteng, Zhu Zhanxing. Spatiotemporal graph convolutional networks: A deep learning framework for traffic forecasting [C] //Proc of the 27th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2018: 3634−3640
39
Guo Shengnan, Lin Youfang, Feng Ning. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting [C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA : AAAI , 2019: 922−929
40
Zheng Chuanpan, Fan Xiaoliang, Wang Cheng, et al. GMAN: A graph multi-attention network for traffic prediction [C] //Proc of the 34th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 1234−1241
41
Ying Rex, He Ruining, Chen Kaifeng, et al. Graph convolutional neural networks for web-scale recommender systems[C] //Proc of the 24th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2018: 974−983
42
Fan Wenqi, Ma Yao, Li Qing, et al. Graph neural networks for social recommendation[C] //Proc of the 19th World Wide Web Conf. New York: ACM, 2019: 417−426
43
Wu Qitian, Zhang Hengrui, Gao Xiaofeng, et al. Dual graph attention networks for deep latent representation of multifaceted social effects in recommender systems[C] //Proc of the 19th World Wide Web. New York: ACM, 2019: 2091−2102
44
Zitnik M, Agrawal M, Leskovec J. Modeling polypharmacy side effects with graph convolutional networks[J]. Bioinformatics, 2018, 34(13): i457−i466
45
Xu Nuo, Wang Pinghui, Chen Long, et al. MR-GNN: Multi- resolution and dual graph neural network for predicting structured entity interactions[C] //Proc of the 28th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2019: 3968−3974
46
Do K, Tran T, Venkatesh S. Graph transformation policy network for chemical reaction prediction [C] //Proc of the 25th ACM SIGKDD Inter Conf on Knowledge Discovery & Data Mining. New York: ACM, 2019: 750−760
47
Schlichtkrull M, Kipf T, Bloem P, et al. Modeling relational data with graph convolutional networks [C] //Proc of the European Semantic Web Conf 2018. Berlin: Springer, 2018: 593−607
48
Chao Shang, Tang Yun, Huang Jing, et al. End-to-end structure-aware convolutional networks for knowledge base completion [C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 3060−3067
49
Zhang Fanjin, Liu Xinyu, Tang Jie, et al, OAG: Toward linking large-scale heterogeneous entity graphs [C]//Proc of the 25th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2019: 2585−2595
50
Bretto A. Hypergraph Theory [M]. Berlin: Springer, 2013: 1−42
51
Zhu Junjie, Zhao Xibin, Hu Han, et al. Emotion recognition from physiological signals using multi-hypergraph neural Networks[C] //Proc of the 2019 IEEE Int Conf on Multimedia and Expo. Piscataway, NJ: IEEE, 2019: 610−615
52
Jiang Jianwen, Wei Yuxuan, Feng Yifan, et al. Dynamic hypergraph neural networks [C] //Proc of the 28th Int Joint Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 2635−2641
53
Shao Jingzhi, Zhu Junjie, Wei Yuxuan, et al. Emotion recognition by edge-weighted hypergraph neural network [C] //Proc of the 2019 IEEE Int Conf on Image Processing. Piscataway, NJ: IEEE, 2019: 2144−2148
54
Fu Sichao, Liu Weifeng, Zhou Yicong et al. HpLapGCN: Hypergraph p-Laplacian graph convolutional networks[J]. Neurocomputing, 2019, 362: 166−174
55
Wang Kesu, Chen Jing, Liao Shijie, et al. Geographic-semantic- temporal hypergraph convolutional network for traffic flow prediction [C] //Proc of the 25th Intl Conf on Pattern Recognition. Piscataway, NJ: IEEE, 2021: 5444−5450
56
Bandyopadhyay S, Das K, Murty M N, Line hypergraph convolution network: Applying graph convolution for hypergraphs [J]. arXiv preprint, arXiv: 2002.03392, 2020
57
Tran L H, Tran L H. Directed hypergraph neural network[J]. arXiv preprint, arXiv: 2008. 03626, 2020
58
Zhang Yubo, Wang Nan, Chen Yufeng, et al. Hypergraph label propagation network[C] // Proc of the 34th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 6885−6892
59
Yang Chaoqi, Wang Ruijie, Yao Shuochao, et al. Hypergraph learning with line expansion[J]. arXiv preprint, arXiv: 2005. 04843, 2020
60
Kim E S, Kang W Y, On K W, et al. Hypergraph attention networks for multimodal learning [C] // Proc of 2020 IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2020: 14569−14578
61
Dong Yihe, Sawin W, Bengio Y. HNHN: Hypergraph networks with hyperedge neurons[J]. arXiv preprint, arXiv: 2006. 12278, 2020
62
Han Jiale, Cheng Bo, Wang Xu. Two-phase hypergraph based reasoning with dynamic relations for multi-hop KBQA [C] //Proc of the 29th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2021: 3615–3621
63
Te Gusi, Hu Wei, Guo Zongming, et al. Exploring hypergraph representation on face anti-spoofing beyond 2D attacks[C] //Proc of the 2020 IEEE Int Conf on Multimedia and Expo. Piscataway, NJ: IEEE , 2020[2021-02-01].https://ieeexplore.ieee.org/document/9102720/
64
Liu Shengyuan, Lv Pei, Zhang Yuzhen, et al. Semi-dynamic hypergraph neural network for 3D pose estimation [C] //Proc of the 29th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2020: 782−788
65
Fatemi B, Taslakian P, Vázquez D, et al. Knowledge hypergraphs: Prediction beyond binary relations [C] //Proc of the 29th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2020: 2191−2197
66
Yi J, Park J. Hypergraph convolutional recurrent neural network [C] //Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2020: 3366−3376
67
Ji Shuyi, Feng Yifan, Ji Rongrong, et al. Dual channel hypergraph collaborative filtering [C] //Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2020: 2020−2029
68
Lostar M, Rekik I. Deep hypergraph U-Net for brain graph embedding and classification[J]. arXiv preprint, arXiv: 2008. 13118, 2020
69
Banka A, Buzi I, Rekik I. Multi-view brain hyperconnectome autoencoder for brain state classification [C] // Proc of the 3rd Int Workshop of Predictive Intelligence in Medicine. Berlin: Springer, 2020: 101−110
70
Yadati N. Neural message passing for multi-relational ordered and recursive hypergraphs [C]//Proc of the 34th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2020: 3275–3289
71
Wu Xiangping, Chen Qingcai, Li Wei, et al. AdaHGNN: Adaptive hypergraph neural networks for multi-Label image classification [C] //Proc of the 28th ACM Inter Conf on Multimedia. New York: ACM, 2020: 284−293
72
Sun Xiangguo, Yin Hongzhi, Liu Bo, et al. Heterogeneous hypergraph embedding for graph classification [C] //Proc of the 14th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2021: 725−733
73
Shao Shuai, Xu Rui, Wang Yanjiang, et al. SAHDL: Sparse attention hypergraph regularized dictionary learning[J]. arXiv preprint, arXiv: 2010.12416, 2020
74
Arya D, Gupta D K, Rudinac S, et al. HyperSAGE: Generalizing inductive representation learning on hypergraphs[J]. arXiv preprint, arXiv: 2010. 04558, 2020
75
Ding Kaize, Wang Jianling, Li Jundong, et al. Be more with less: Hypergraph attention networks for inductive text classification [C] //Proc of the 2020 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2020: 4927−4936
76
Sawhney R, Agarwal S, Wadhwa A, et al. Spatiotemporal hypergraph convolution network for stock movement forecasting [C] //Proc of the 2020 IEEE Int Conf on Data Mining. Piscataway, NJ: IEEE, 2020: 482−491
77
Tan Yaqi, Ma Zhongchen, Zhan Yongzhao, et al. Hypergraph induced graph convolutional network for multi-label image recognition[C/OL] //Proc of the 2020 Int Conf on Internet of Things and Intelligent Applications. Piscataway, NJ: IEEE, 2020[2021-03-01].https://ieeexplore.ieee.org/document/9312371
78
Hou Rui, Small M, Forrest A R R. Community detection in a weighted directed hypergraph representation of cell-to-cell communication networks [J/OL]. bioRxiv, 2020[2021-03-01].https://doi.org/10.1101/ 2020.11.16.381566
79
Bandyopadhyay S, Das K, Murty M N, Hypergraph attention isomorphism network by learning line graph expansion [C] //Proc of the 2020 IEEE Int Conf on Big Data. Piscataway, NJ: IEEE, 2020: 669−678
80
Garasuie M M, Shabankhah M, Kamandi A. Improving hypergraph attention and hypergraph convolution networks [C] //Proc of the 11th Int Conf on Information and Knowledge Technology. Piscataway, NJ: IEEE, 2020: 67−72
81
Xia Xin, Yin Hongzhi, Yu Junliang, et al. Self-supervised hypergraph convolutional networks for session-based recommendation [C]// Proc of the 35th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2021: 4503−4511
82
Sawhney R, Agarwal S, Wadhwa A, et al. Stock selection via spatiotemporal hypergraph attention network: A learning to rank approach [C] //Proc of the 35th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2021: 497−504
83
Yu Junliang, Yin Hongzhi, Li Jundong, et al. Self-supervised multi-channel hypergraph convolutional network for social recommendation [C] //Proc of the 21st World Wide Web. New York: ACM, 2021: 413−424
84
Hao Xiaoke, Li Jie, Yingchun, et al. Hypergraph neural network for skeleton-based action recognition[J]. IEEE Transactions on Image Processing, 2021, 30: 2263−2275
85
Xue Hansheng, Yang Luwei, Rajan V, et al. Multiplex bipartite network embedding using dual hypergraph convolutional networks [C] //Proc of the 21st World Wide Web. New York: ACM, 2021: 1649−1660
86
Maleki S, Wall D P, Pingali K. NetVec: A dcalable hypergraph embedding system[J]. arXiv preprint, arXiv: 2103. 09660v1, 2021
87
Tudisco F, Prokopchik K, Benson A R. A nonlinear diffusion method for semi-supervised learning on hypergraphs[J]. arXiv preprint, arXiv: 2103. 14867, 2021
88
Wang Jingcheng, Zhang Yong, Wei Yun, et al. Metro passenger flow prediction via dynamic hypergraph convolution networks[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(12): 7891−7903
89
Liu Binghao, Zhao Pengpeng, Zhuang Fuzhen, et al. Knowledge-aware hypergraph neural network for recommender systems[C] //Proc of the 26th Int Conf on Database Systems for Advanced Applications. Berlin: Springer, 2021: 132−147
90
Wang Jianling, Ding Kaize, Zhu Ziwei, et al. Session-based recommendation with hypergraph attention networks[J]. arXiv preprint, arXiv: 2112. 14266, 2021
91
Bai Junjie, Gong Biao, Zhao Yining, et al. Multi-scale representation learning on hypergraph for 3D shape retrieval and recognition[J]. IEEE Transactions on Image Processing, 2021, 30: 5327−5338
92
Du Boxin, Yuan Changhe, Robert B, et al. Hypergraph pre-training with graph neural networks[J]. arXiv preprint, arXiv: 2105. 10862, 2021
93
Huang Jing, Huang Xiaolin, Yang Jie. Residual enhanced multi- hypergraph neural network [C] // Proc of the 2021 IEEE Int Conf on Image Processing. Piscataway, NJ: IEEE, 2021: 3657−3661
94
Fu Jun, Hou Chengbin, Zhou Wei, et al. Adaptive hypergraph convolutional network for no-reference 360-degree image quality assessment [C] //Proc of the 30th ACM Inte Conf on Multimedia. New York: ACM, 2021: 961–969
95
Huang Jing, Yang Jie. UniGNN: A unified framework for graph and hypergraph neural networks [C] //Proc of the 30th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2021: 2563−2569
96
Jo J, Baek Ji, Lee S, et al. Edge representation learning with hypergraphs[J]. arXiv preprint, arXiv: 2106. 15845, 2021
97
Zhang Jiying, Chen Yuzhao, Xiao Xiong, et al. Learnable hypergraph Laplacian for hypergraph learning [C] //Proc of the 2022 IEEE Int Conf on Acoustics, Speech and Signal Processing. Piscataway, NJ: IEEE, 2022: 4503−4507
98
Yan Yichao, Qin Jie, Chen Jiaxin, et al. Learning multi-granular hypergraphs for video-based person re-identification [C] //Proc of the 2020 IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2020: 2896−2905
99
Chien E, Pan Chao, Peng Jianhao, et al. You are AllSet: A multiset function framework for hypergraph neural networks[J]. arXiv preprint, arXiv: 2106. 13264, 2021
100
Cui Chaoran, Li Xiaojie, Du Juan, et al. Temporal-relational hypergraph tri-attention networks for stock trend prediction[J]. arXiv preprint, arXiv: 2107. 14033, 2021
101
Pan Junren, Lei Baiying, Shen Yanyan, et al. Characterization multimodal connectivity of brain network by hypergraph GAN for Alzheimer’s disease analysis [C] //Proc of the 4th Chinese Conf on Pattern Recognition and Computer Vision. Berlin: Springer, 2021: 467−478
102
Zuo Qinkun, Lei Baiying, Shen Yanyan, et al. Multimodal representations learning and adversarial hypergraph fusion for early Alzheimer’s disease prediction [C] //Proc of the 4th Chinese Conf on Pattern Recognition and Computer Vision. Berlin: Springer, 2021: 479− 490
103
Li Yichao, Chen Hongxu, Sun Xiangguo, et al. Hyperbolic hypergraphs for sequential recommendation [C] //Proc of the 30th ACM Int Conf on Information & Knowledge Management. New York: ACM, 2021: 988−997
104
Yadati N, Gao T, Asoodeh S, et al. Graph neural networks for soft semi-supervised learning on hypergraphs [C] //Proc of the 25th Pacific-Asia Conf on Knowledge Discovery and Data Mining. Berlin: Springer, 2021: 447−458
105
Ma Zhongtian, Jiang Zhiguo, Zhang Haopeng. Hyperspectral image classification using spectral-spatial hypergraph convolution neural network[J]. SPIE Remote Sensing, 2021, 11862: 118620I
106
Wu Longcan, Wang Daling, Song Kaisong, et al. Dual-view hypergraph neural networks for attributed graph learning[J]. Knowledge-Based Systems, 2021, 227: 107185
107
Zhang Junwei, Gao Min, Yu Junliang, et al. Double-scale self- supervised hypergraph learning for group recommendation[C] //Proc of the 30th ACM Int Conf on Information & Knowledge Management. New York: ACM, 2021: 2557−2567
108
Vijaikumar M, Hada D V, Shevade S K. HyperTeNet: Hypergraph and transformer-based neural network for personalized list continuation [C] //Proc of the 2021 IEEE Int Conf on Data Mining. Piscataway, NJ: IEEE 2021: 1210−1215
109
Xia Zhongxiu, Zhang Weiyu, Weng Ziqiang. Social recommendation system based on hypergraph attention network[J]. Computational Intelligence and Neuroscience, 2021, 2021: 7716214
110
Ding Meirong, Lin Xiaokang, Zeng Biqing, et al. Hypergraph neural networks with attention mechanism for session-based recommendation[J]. Journal of Physics: Conference Series, 2021, 2082(1): 012007
111
Zhu Zirui, Gao Chen, Chen Xu, et al. Inhomogeneous social recommendation with hypergraph convolutional networks[J]. arXiv preprint, arXiv: 2111. 03344, 2021
112
Luo Xiaoyi, Peng Jiaheng, Liang Jun. Directed hypergraph attention network for traffic forecasting[J]. IET Intelligent Transport Systems, 2022, 16(1): 85−98
113
Lin Jingjing, Ye Zhonglin, Zhao Haixing, et al. DeepHGNN: A novel deep hypergraph neural network[J]. Chinese Journal of Electronics, 2022, 31(5): 958−968
114
Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks [J]. arXiv preprint, arXiv:1609. 02907, 2017
115
Bruna J, Zaremba W, Szlam A, et al. Spectral networks and locally connected networks on graphs[J]. arXiv preprint, arXiv: 1312. 6203, 2014
116
Defferrard M, Bresson X, Vandergheynst P. Convolutional neural networks on graphs with fast localized spectral filtering[C] //Proc of the 30th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2016: 3844−3852
117
Zhou Dengyong, Huang Jiayuan, Schölkopf B. Learning with hypergraphs: Clustering, classification, and embedding [C] // Proc of the 19th Int Conf on Neural Information Processing Systems. Cambridge, MA: MIT, 2006: 1601–1608
118
Song Bai, Zhang Feihu, Torr P H S. Hypergraph convolution and hypergraph attention[J]. Pattern Recognition, 2021, 110: 107637
119
Ma Xueqi, Liu Weifeng, Li Shuying, et al. Hypergraph p-Laplacian regularization for remotely sensed image recognition[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(3): 1585−1595
120
Donnat C, Zitnik M, Hallac D, et al. Learning structural node embeddings via diffusion wavelets[J]. arXiv preprint, arXiv: 1710. 10321, 2018
121
Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift [C] //Proc of the 32nd Inte Conf on Machine Learning. New York: ACM, 2015: 448−456
122
Louis A. Hypergraph Markov operators, eigenvalues and approximation algorithms [C] //Proc of the 47th Annual ACM Symp on Theory of Computing. New York: ACM, 2015: 713−722
123
Chan T H H, Louis A, Tang Z G, et al. Spectral properties of hypergraph Laplacian and approximation algorithms[J]. Journal of the ACM, 2018, 65(3): 1−48
124
Chan T H H, Liang Zhibin. Generalizing the hypergraph Laplacian via a diffusion process with mediators [C] // Proc of the 24th Int Computing and Combinatorics Computing and Combinatorics. Berlin: Springer, 2020: 441–453
125
Xu Keyulu, Hu Weihua, Leskovec J, et al. How powerful are graph neural networks?[J]. arXiv preprint, arXiv: 1810. 00826, 2019
126
Sun Xiangguo, Yin Hongzhi, Liu Bo, et al. Multi-level hyperedge distillation for social linking prediction on sparsely observed networks[C] //Proc of the 21st World Wide Web Conf. New York: ACM, 2021: 2934−2945
127
Xia Liqiao, Zheng Pai, Huang Xiao, et al. A novel hypergraph convolution network-based approach for predicting the material removal rate in chemical mechanical planarization[J]. Journal of Intelligent Manufacturing, 2021, 33(8): 2295−2306
128
Alfke D, Stoll M. Semi-supervised classification on non-sparse graphs using low-rank graph convolutional networks[J]. arXiv preprint, arXiv: 1905. 10224v1, 2019
129
Chami I, Rex Y, Ré C, et al. Hyperbolic graph convolutional neural networks [J]. arXiv preprint, arXiv : 1910.12933, 2019
130
Srinivasan B, Zheng Da, Karypis G. Learning over families of sets-hypergraph representation learning for higher order tasks[J]. arXiv preprint, arXiv: 2101. 07773v1, 2021
131
Chen Ming, Wei Zhewei, Huang Zengfeng, et al. Simple and deep graph convolutional networks[C] //Proc of the 37th Int Conf on Machine Learning. New York: ACM, 2020: 1725−1735
132
Payne J. Deep hyperedges: A framework for transductive and inductive learning on hypergraphs[J]. arXiv preprint, arXiv: 1910. 02633, 2019
133
Gilmer J, Schoenholz S S, Riley P F, et al. Neural message passing for quantum chemistry [C] //Proc of the 34th Int Conf on Machine Learning. New York: ACM, 2017: 1263−1272
134
Hamilton W L, Ying R, Jure L. Inductive representation learning on large graphs[C] //Proc of the 31st Int Conf on Neural Information Processing Systems. Red Hook: Curran Associates Inc, 2017: 1025−1035
135
Liu Qiao, Zeng Yifu, Mokhosi R, et al. STAMP: Short-term attention/memory priority model for session-based recommendation [C] //Proc of the 24th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2018: 1831−1839
136
Wu Shu, Tang Yuyuan, Zhu Yanqiao, et al. Session-based recommendation with graph neural networks[C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2019: 346–353
137
He Xiangnan, Deng Kuan, Wang Xiang, et al. LightGCN: Simplifying and powering graph convolution network for recommendation [C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 639−648
138
Wadhwa G, Dhall A, Murala S, et al. Hyperrealistic image inpainting with hypergraphs[C] //Proc of the 2021 IEEE Winter Conf on Applications of Computer Vision. Piscataway, NJ: IEEE, 2021: 3911−3920
139
Dang N T V, Tran L H, Tran L H. Noise-robust classification with hypergraph neural network[J]. arXiv preprint, arXiv: 2102. 01934, 2021
140
Zheng Wenbo, Yan Lan, Gou Chao, et al. Two heads are better than one: Hypergraph-enhanced graph reasoning for visual event ratiocination [C] //Proc of the 38th Int Conf on Machine Learning. New York: ACM, 139: 12747−12760
141
Zhao Weizhong, Zhang Jinyong, Yang Jincai, et al. A novel joint biomedical event extraction framework via two-level modeling of documents[J]. Information Sciences, 2021, 550: 27−40
142
Vashishth S, Sanyal S, Nitin V, et al. Composition-based multi-relational graph convolutional networks[J]. arXiv preprint, arXiv: 1911. 038082, 2020
143
Yadati N, Dayanidhi R, Vaishnavi S, et al. Knowledge base question answering through recursive hypergraphs[C] //Proc of the 16th Conf of the European Chapter of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2021: 448−454
144
Han Jiale, Cheng Bo, Wang Xu. Open domain question answering based on text enhanced knowledge graph with hyperedge infusion [C] //Proc of the 2020 Findings of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2020: 1475−1481
145
Miller A, Fisch A, Dodge J, et al. Key-value memory networks for directly reading documents [C] //Proc of the 2016 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2016: 1400–1409
146
Sun Haitian, Dhingra B, Zaheer M, et al. Open domain question answering using early fusion of knowledge bases and text[J]. arXiv preprint, arXiv: 1809, 00782, 2018
147
Dongen S V. Graph clustering by flow simulation [D]. Amsterdam: Center for Math and Computer Science, 2000
148
Blondel V D, Guillaume J L, Lambiotte R, et al. Fast unfolding of communities in large networks[J]. Journal of Statistical Mechanics: Theory and Experiment, 2008, 2008(10): 10008
149
Madine M M, Rekik I, Werghi N. Diagnosing autism using T1-W MRI with multi-kernel learning and hypergraph neural network [C] //Proc of the 2020 IEEE Int Conf on Image Processing. Piscataway, NJ: IEEE, 2020: 438−442
150
Xu Keyulu, Li Chengtao, Tian Yonglong, et al. Representation learning on graphs with jumping knowledge networks[C] //Proc of the 35th Int Conf on Machine Learning. New York: ACM, 2018: 5453−5462
151
Yu Rong, Huang Wenbing, Xu Tingyang, et al. DropEdge: Towards deep graph convolutional networks on node classification[J]. arXiv preprint, arXiv: 1907. 10903, 2020
2024年第61卷第2期
94
45
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202220483
  • 接收时间:2022-06-10
  • 首发时间:2026-11-17
  • 出版时间:2024-02-02
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2022-06-10
  • 修回日期:2023-04-18
基金
作者信息
    1青海师范大学计算机学院 西宁 810008
    2藏语智能信息处理及应用国家重点实验室(青海师范大学) 西宁 810008
    3西宁城市职业技术学院信息工程系 西宁 810003

通讯作者:

赵海兴()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202220483
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏