We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Social and Information Networks

Title: VR-GNN: Variational Relation Vector Graph Neural Network for Modeling both Homophily and Heterophily

Abstract: Graph Neural Networks (GNNs) have achieved remarkable success in diverse real-world applications. Traditional GNNs are designed based on homophily, which leads to poor performance under heterophily scenarios. Current solutions deal with heterophily mainly by mixing high-order neighbors or passing signed messages. However, mixing high-order neighbors destroys the original graph structure and passing signed messages utilizes an inflexible message-passing mechanism, which is prone to producing unsatisfactory effects. To overcome the above problems, we propose a novel GNN model based on relation vector translation named Variational Relation Vector Graph Neural Network (VR-GNN). VR-GNN models relation generation and graph aggregation into an end-to-end model based on Variational Auto-Encoder. The encoder utilizes the structure, feature and label to generate a proper relation vector. The decoder achieves superior node representation by incorporating the relation translation into the message-passing framework. VR-GNN can fully capture the homophily and heterophily between nodes due to the great flexibility of relation translation in modeling neighbor relationships. We conduct extensive experiments on eight real-world datasets with different homophily-heterophily properties to verify the effectiveness of our model. The experimental results show that VR-GNN gains consistent and significant improvements against state-of-the-art GNN methods under heterophily, and competitive performance under homophily.
Subjects: Social and Information Networks (cs.SI)
Cite as: arXiv:2211.14523 [cs.SI]
  (or arXiv:2211.14523v4 [cs.SI] for this version)

Submission history

From: Ren Li [view email]
[v1] Sat, 26 Nov 2022 09:21:15 GMT (2773kb,D)
[v2] Fri, 2 Dec 2022 15:28:46 GMT (5544kb,D)
[v3] Fri, 20 Jan 2023 17:32:28 GMT (3267kb,D)
[v4] Tue, 24 Jan 2023 19:02:56 GMT (3267kb,D)

Link back to: arXiv, form interface, contact.