Graph neural convection-diffusion with heterophily
Graph neural networks (GNNs) have shown promising results across various graph learning tasks, but they often assume homophily, which can result in poor performance on heterophilic graphs. The connected nodes are likely to be from different classes or have dissimilar features on heterophilic grap...
| Main Authors: | , , , , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/171667 https://www.ijcai.org/proceedings/2023/ |
| _version_ | 1826126818790342656 |
|---|---|
| author | Zhao, Kai Kang, Qiyu Song, Yang She, Rui Wang, Sijie Tay, Wee Peng |
| author2 | School of Electrical and Electronic Engineering |
| author_facet | School of Electrical and Electronic Engineering Zhao, Kai Kang, Qiyu Song, Yang She, Rui Wang, Sijie Tay, Wee Peng |
| author_sort | Zhao, Kai |
| collection | NTU |
| description | Graph neural networks (GNNs) have shown promising results across various
graph learning tasks, but they often assume homophily, which can result in poor
performance on heterophilic graphs. The connected nodes are likely to be from
different classes or have dissimilar features on heterophilic graphs. In this
paper, we propose a novel GNN that incorporates the principle of heterophily by
modeling the flow of information on nodes using the convection-diffusion
equation (CDE). This allows the CDE to take into account both the diffusion of
information due to homophily and the ``convection'' of information due to
heterophily. We conduct extensive experiments, which suggest that our framework
can achieve competitive performance on node classification tasks for
heterophilic graphs, compared to the state-of-the-art methods. The code is
available at \url{https://github.com/zknus/Graph-Diffusion-CDE}. |
| first_indexed | 2024-10-01T06:58:34Z |
| format | Conference Paper |
| id | ntu-10356/171667 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T06:58:34Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | ntu-10356/1716672023-12-01T15:39:15Z Graph neural convection-diffusion with heterophily Zhao, Kai Kang, Qiyu Song, Yang She, Rui Wang, Sijie Tay, Wee Peng School of Electrical and Electronic Engineering Thirty-Second International Joint Conference on Artificial Intelligence (IJCAI-23) Centre for Information Sciences and Systems (CISS) Engineering::Electrical and electronic engineering Machine Learning Classification Graph neural networks (GNNs) have shown promising results across various graph learning tasks, but they often assume homophily, which can result in poor performance on heterophilic graphs. The connected nodes are likely to be from different classes or have dissimilar features on heterophilic graphs. In this paper, we propose a novel GNN that incorporates the principle of heterophily by modeling the flow of information on nodes using the convection-diffusion equation (CDE). This allows the CDE to take into account both the diffusion of information due to homophily and the ``convection'' of information due to heterophily. We conduct extensive experiments, which suggest that our framework can achieve competitive performance on node classification tasks for heterophilic graphs, compared to the state-of-the-art methods. The code is available at \url{https://github.com/zknus/Graph-Diffusion-CDE}. Agency for Science, Technology and Research (A*STAR) Published version This research is supported by A*STAR under its RIE2020 Advanced Manufacturing and Engineering (AME) Industry Alignment Fund – Pre Positioning (IAF-PP) (Grant No. A19D6a0053) and the National Research Foundation, Singapore and Infocomm Media Development Authority under its Future Communications Research and Development Programme. The computational work for this article was (fully/partially) performed on resources of the National Supercomputing Centre, Singapore (https://www.nscc.sg). 2023-11-27T02:21:19Z 2023-11-27T02:21:19Z 2023 Conference Paper Zhao, K., Kang, Q., Song, Y., She, R., Wang, S. & Tay, W. P. (2023). Graph neural convection-diffusion with heterophily. Thirty-Second International Joint Conference on Artificial Intelligence (IJCAI-23), 4656-4664. https://dx.doi.org/10.24963/ijcai.2023/518 978-1-956792-03-4 https://hdl.handle.net/10356/171667 10.24963/ijcai.2023/518 https://www.ijcai.org/proceedings/2023/ 4656 4664 en A19D6a0053 © 2023 International Joint Conferences on Artificial Intelligence. All rights reserved.This article may be downloaded for personal use only. Any other use requires prior permission of the copyright holder. The Version of Record is available online at http://doi.org/10.24963/ijcai.2023/518. application/pdf |
| spellingShingle | Engineering::Electrical and electronic engineering Machine Learning Classification Zhao, Kai Kang, Qiyu Song, Yang She, Rui Wang, Sijie Tay, Wee Peng Graph neural convection-diffusion with heterophily |
| title | Graph neural convection-diffusion with heterophily |
| title_full | Graph neural convection-diffusion with heterophily |
| title_fullStr | Graph neural convection-diffusion with heterophily |
| title_full_unstemmed | Graph neural convection-diffusion with heterophily |
| title_short | Graph neural convection-diffusion with heterophily |
| title_sort | graph neural convection diffusion with heterophily |
| topic | Engineering::Electrical and electronic engineering Machine Learning Classification |
| url | https://hdl.handle.net/10356/171667 https://www.ijcai.org/proceedings/2023/ |
| work_keys_str_mv | AT zhaokai graphneuralconvectiondiffusionwithheterophily AT kangqiyu graphneuralconvectiondiffusionwithheterophily AT songyang graphneuralconvectiondiffusionwithheterophily AT sherui graphneuralconvectiondiffusionwithheterophily AT wangsijie graphneuralconvectiondiffusionwithheterophily AT tayweepeng graphneuralconvectiondiffusionwithheterophily |