A Transformative Topological Representation for Link Modeling, Prediction and Cross-Domain Network Analysis

被引:0
作者
Zhang, Kai [1 ]
Shen, Junchen [1 ]
He, Gaoqi [3 ]
Sun, Yu [2 ]
Ling, Haibin [3 ]
Zha, Hongyuan [4 ]
Li, Honglin [5 ,6 ]
Zhang, Jie [7 ]
机构
[1] East China Normal Univ, Sch Comp Sci & Technol, Shanghai, Peoples R China
[2] Indeed Inc, Sunnyvale, CA 94086 USA
[3] SUNY Stony Brook Univ, Dept Comp Sci, Stony Brook, NY USA
[4] Chinese Univ Hong Kong, Sch Data Sci, Shenzhen 518172, Guangdong, Peoples R China
[5] East China Normal Univ, Innovat Ctr AI & Drug Discovery, Shanghai 200062, Peoples R China
[6] Lingang Lab, Shanghai 200031, Peoples R China
[7] Fudan Univ, Inst Sci & Technol Brain inspired Intelligence, Shanghai 200438, Peoples R China
基金
中国国家自然科学基金;
关键词
Predictive models; Feature extraction; Vectors; Topology; Graph neural networks; Prediction algorithms; Kernel; Complex networks; graph neural networks; link prediction; topological representation; COMPLEX NETWORKS; CLASSIFICATION; SYSTEMS;
D O I
10.1109/TPAMI.2024.3378729
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Many complex social, biological, or physical systems are characterized as networks, and recovering the missing links of a network could shed important lights on its structure and dynamics. A good topological representation is crucial to accurate link modeling and prediction, yet how to account for the kaleidoscopic changes in link formation patterns remains a challenge, especially for analysis in cross-domain studies. We propose a new link representation scheme by projecting the local environment of a link into a "dipole plane", where neighboring nodes of the link are positioned via their relative proximity to the two anchors of the link, like a dipole. By doing this, complex and discrete topology arising from link formation is turned to differentiable point-cloud distribution, opening up new possibilities for topological feature-engineering with desired expressiveness, interpretability and generalization. Our approach has comparable or even superior results against state-of-the-art GNNs, meanwhile with a model up to hundreds of times smaller and running much faster. Furthermore, it provides a universal platform to systematically profile, study, and compare link-patterns from miscellaneous real-world networks. This allows building a global link-pattern atlas, based on which we have uncovered interesting common patterns of link formation, i.e., the bridge-style, the radiation-style, and the community-style across a wide collection of networks with highly different nature.
引用
收藏
页码:6126 / 6138
页数:13
相关论文
共 60 条
[1]   Friends and neighbors on the Web [J].
Adamic, LA ;
Adar, E .
SOCIAL NETWORKS, 2003, 25 (03) :211-230
[2]   Statistical mechanics of complex networks [J].
Albert, R ;
Barabási, AL .
REVIEWS OF MODERN PHYSICS, 2002, 74 (01) :47-97
[3]   Classes of small-world networks [J].
Amaral, LAN ;
Scala, A ;
Barthélémy, M ;
Stanley, HE .
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA, 2000, 97 (21) :11149-11152
[4]   REFINED STATUS INDEX FOR SOCIOMETRIC DATA [J].
ARNEY, WR .
SOCIOLOGICAL METHODS & RESEARCH, 1973, 1 (03) :329-345
[5]  
Barabasi AL, 2016, NETWORK SCIENCE, P1
[6]   Emergence of scaling in random networks [J].
Barabási, AL ;
Albert, R .
SCIENCE, 1999, 286 (5439) :509-512
[7]  
Batagelj V., 2006, PAJEK DATASETS
[8]   Representation Learning: A Review and New Perspectives [J].
Bengio, Yoshua ;
Courville, Aaron ;
Vincent, Pascal .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2013, 35 (08) :1798-1828
[9]   Complex networks: Structure and dynamics [J].
Boccaletti, S. ;
Latora, V. ;
Moreno, Y. ;
Chavez, M. ;
Hwang, D. -U. .
PHYSICS REPORTS-REVIEW SECTION OF PHYSICS LETTERS, 2006, 424 (4-5) :175-308
[10]   The anatomy of a large-scale hypertextual Web search engine [J].
Brin, S ;
Page, L .
COMPUTER NETWORKS AND ISDN SYSTEMS, 1998, 30 (1-7) :107-117