File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: SP-GNN: Learning structure and position information from graphs

TitleSP-GNN: Learning structure and position information from graphs
Authors
KeywordsGraph classification
Graph neural networks
Node classification
Positional embedding
Structural embedding
Issue Date15-Apr-2023
PublisherElsevier
Citation
Neural Networks, 2023, v. 161, p. 505-514 How to Cite?
Abstract

Graph neural network (GNN) is a powerful model for learning from graph data. However, existing GNNs may have limited expressive power, especially in terms of capturing adequate structural and positional information of input graphs. Structure properties and node position information are unique to graph-structured data, but few GNNs are capable of capturing them. This paper proposes Structure- and Position-aware Graph Neural Networks (SP-GNN), a new class of GNNs offering generic and expressive power of graph data. SP-GNN enhances the expressive power of GNN architectures by incorporating a near-isometric proximity-aware position encoder and a scalable structure encoder. Further, given a GNN learning task, SP-GNN can be used to analyze positional and structural awareness of GNN tasks using the corresponding embeddings computed by the encoders. The awareness scores can guide fusion strategies of the extracted positional and structural information with raw features for better performance of GNNs on downstream tasks. We conduct extensive experiments using SP-GNN on various graph datasets and observe significant improvement in classification over existing GNN models.


Persistent Identifierhttp://hdl.handle.net/10722/331957
ISSN
2023 Impact Factor: 6.0
2023 SCImago Journal Rankings: 2.605
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorChen, Yangrui-
dc.contributor.authorYou, Jiaxuan-
dc.contributor.authorHe, Jun-
dc.contributor.authorLin, Yuan-
dc.contributor.authorPeng, Yanghua-
dc.contributor.authorWu, Chuan-
dc.contributor.authorZhu, Yibo-
dc.date.accessioned2023-09-28T04:59:52Z-
dc.date.available2023-09-28T04:59:52Z-
dc.date.issued2023-04-15-
dc.identifier.citationNeural Networks, 2023, v. 161, p. 505-514-
dc.identifier.issn0893-6080-
dc.identifier.urihttp://hdl.handle.net/10722/331957-
dc.description.abstract<p>Graph <a href="https://www.sciencedirect.com/topics/mathematics/neural-network" title="Learn more about neural network from ScienceDirect's AI-generated Topic Pages">neural network</a> (GNN) is a powerful model for learning from graph data. However, existing GNNs may have limited <a href="https://www.sciencedirect.com/topics/computer-science/expressive-power" title="Learn more about expressive power from ScienceDirect's AI-generated Topic Pages">expressive power</a>, especially in terms of capturing adequate structural and positional information of input graphs. Structure properties and node position information are unique to graph-structured data, but few GNNs are capable of capturing them. This paper proposes <em>Structure-</em> and <em>Position-aware Graph Neural Networks (SP-GNN)</em>, a new class of GNNs offering generic and <a href="https://www.sciencedirect.com/topics/computer-science/expressive-power" title="Learn more about expressive power from ScienceDirect's AI-generated Topic Pages">expressive power</a> of graph data. SP-GNN enhances the expressive power of GNN architectures by incorporating a near-isometric proximity-aware position encoder and a scalable structure encoder. Further, given a GNN learning task, SP-GNN can be used to analyze positional and structural awareness of GNN tasks using the corresponding embeddings computed by the encoders. The awareness scores can guide fusion strategies of the extracted positional and structural information with raw features for better performance of GNNs on downstream tasks. We conduct extensive experiments using SP-GNN on various graph datasets and observe significant improvement in classification over existing GNN models.<br></p>-
dc.languageeng-
dc.publisherElsevier-
dc.relation.ispartofNeural Networks-
dc.subjectGraph classification-
dc.subjectGraph neural networks-
dc.subjectNode classification-
dc.subjectPositional embedding-
dc.subjectStructural embedding-
dc.titleSP-GNN: Learning structure and position information from graphs-
dc.typeArticle-
dc.identifier.doi10.1016/j.neunet.2023.01.051-
dc.identifier.scopuseid_2-s2.0-85148326493-
dc.identifier.volume161-
dc.identifier.spage505-
dc.identifier.epage514-
dc.identifier.eissn1879-2782-
dc.identifier.isiWOS:000944692100001-
dc.identifier.issnl0893-6080-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats