LGNN: a novel linear graph neural network algorithm
The emergence of deep learning has not only brought great changes in the field of image recognition, but also achieved excellent node classification performance in graph neural networks. However, the existing graph neural network framework often uses methods based on spatial domain or spectral domai...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2023-11-01
|
| Series: | Frontiers in Computational Neuroscience |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fncom.2023.1288842/full |
| _version_ | 1797400568599150592 |
|---|---|
| author | Shujuan Cao Shujuan Cao Shujuan Cao Shujuan Cao Xiaoming Wang Zhonglin Ye Zhonglin Ye Zhonglin Ye Zhonglin Ye Mingyuan Li Mingyuan Li Mingyuan Li Mingyuan Li Haixing Zhao Haixing Zhao Haixing Zhao Haixing Zhao |
| author_facet | Shujuan Cao Shujuan Cao Shujuan Cao Shujuan Cao Xiaoming Wang Zhonglin Ye Zhonglin Ye Zhonglin Ye Zhonglin Ye Mingyuan Li Mingyuan Li Mingyuan Li Mingyuan Li Haixing Zhao Haixing Zhao Haixing Zhao Haixing Zhao |
| author_sort | Shujuan Cao |
| collection | DOAJ |
| description | The emergence of deep learning has not only brought great changes in the field of image recognition, but also achieved excellent node classification performance in graph neural networks. However, the existing graph neural network framework often uses methods based on spatial domain or spectral domain to capture network structure features. This process captures the local structural characteristics of graph data, and the convolution process has a large amount of calculation. It is necessary to use multi-channel or deep neural network structure to achieve the goal of modeling the high-order structural characteristics of the network. Therefore, this paper proposes a linear graph neural network framework [Linear Graph Neural Network (LGNN)] with superior performance. The model first preprocesses the input graph, and uses symmetric normalization and feature normalization to remove deviations in the structure and features. Then, by designing a high-order adjacency matrix propagation mechanism, LGNN enables nodes to iteratively aggregate and learn the feature information of high-order neighbors. After obtaining the node representation of the network structure, LGNN uses a simple linear mapping to maintain computational efficiency and obtain the final node representation. The experimental results show that the performance of the LGNN algorithm in some tasks is slightly worse than that of the existing mainstream graph neural network algorithms, but it shows or exceeds the machine learning performance of the existing algorithms in most graph neural network performance evaluation tasks, especially on sparse networks. |
| first_indexed | 2024-03-09T01:57:26Z |
| format | Article |
| id | doaj.art-1cdccb1c4a474c4c832bb829f4ff2075 |
| institution | Directory Open Access Journal |
| issn | 1662-5188 |
| language | English |
| last_indexed | 2024-03-09T01:57:26Z |
| publishDate | 2023-11-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Computational Neuroscience |
| spelling | doaj.art-1cdccb1c4a474c4c832bb829f4ff20752023-12-08T13:00:08ZengFrontiers Media S.A.Frontiers in Computational Neuroscience1662-51882023-11-011710.3389/fncom.2023.12888421288842LGNN: a novel linear graph neural network algorithmShujuan Cao0Shujuan Cao1Shujuan Cao2Shujuan Cao3Xiaoming Wang4Zhonglin Ye5Zhonglin Ye6Zhonglin Ye7Zhonglin Ye8Mingyuan Li9Mingyuan Li10Mingyuan Li11Mingyuan Li12Haixing Zhao13Haixing Zhao14Haixing Zhao15Haixing Zhao16College of Computer, Qinghai Normal University, Xining, Qinghai, ChinaSchool of Computer Science, Shaanxi Normal University, Xi’an, Shaanxi, ChinaThe State Key Laboratory of Tibetan Intelligent Information Processing and Application, Xining, Qinghai, ChinaKey Laboratory of Tibetan Information Processing, Ministry of Education, Xining, Qinghai, ChinaSchool of Computer Science, Shaanxi Normal University, Xi’an, Shaanxi, ChinaCollege of Computer, Qinghai Normal University, Xining, Qinghai, ChinaSchool of Computer Science, Shaanxi Normal University, Xi’an, Shaanxi, ChinaThe State Key Laboratory of Tibetan Intelligent Information Processing and Application, Xining, Qinghai, ChinaKey Laboratory of Tibetan Information Processing, Ministry of Education, Xining, Qinghai, ChinaCollege of Computer, Qinghai Normal University, Xining, Qinghai, ChinaSchool of Computer Science, Shaanxi Normal University, Xi’an, Shaanxi, ChinaThe State Key Laboratory of Tibetan Intelligent Information Processing and Application, Xining, Qinghai, ChinaKey Laboratory of Tibetan Information Processing, Ministry of Education, Xining, Qinghai, ChinaCollege of Computer, Qinghai Normal University, Xining, Qinghai, ChinaSchool of Computer Science, Shaanxi Normal University, Xi’an, Shaanxi, ChinaThe State Key Laboratory of Tibetan Intelligent Information Processing and Application, Xining, Qinghai, ChinaKey Laboratory of Tibetan Information Processing, Ministry of Education, Xining, Qinghai, ChinaThe emergence of deep learning has not only brought great changes in the field of image recognition, but also achieved excellent node classification performance in graph neural networks. However, the existing graph neural network framework often uses methods based on spatial domain or spectral domain to capture network structure features. This process captures the local structural characteristics of graph data, and the convolution process has a large amount of calculation. It is necessary to use multi-channel or deep neural network structure to achieve the goal of modeling the high-order structural characteristics of the network. Therefore, this paper proposes a linear graph neural network framework [Linear Graph Neural Network (LGNN)] with superior performance. The model first preprocesses the input graph, and uses symmetric normalization and feature normalization to remove deviations in the structure and features. Then, by designing a high-order adjacency matrix propagation mechanism, LGNN enables nodes to iteratively aggregate and learn the feature information of high-order neighbors. After obtaining the node representation of the network structure, LGNN uses a simple linear mapping to maintain computational efficiency and obtain the final node representation. The experimental results show that the performance of the LGNN algorithm in some tasks is slightly worse than that of the existing mainstream graph neural network algorithms, but it shows or exceeds the machine learning performance of the existing algorithms in most graph neural network performance evaluation tasks, especially on sparse networks.https://www.frontiersin.org/articles/10.3389/fncom.2023.1288842/fullgraph neural networklinear neural networkgraph deep learninggraph representation learninghigh-order structural constraint |
| spellingShingle | Shujuan Cao Shujuan Cao Shujuan Cao Shujuan Cao Xiaoming Wang Zhonglin Ye Zhonglin Ye Zhonglin Ye Zhonglin Ye Mingyuan Li Mingyuan Li Mingyuan Li Mingyuan Li Haixing Zhao Haixing Zhao Haixing Zhao Haixing Zhao LGNN: a novel linear graph neural network algorithm Frontiers in Computational Neuroscience graph neural network linear neural network graph deep learning graph representation learning high-order structural constraint |
| title | LGNN: a novel linear graph neural network algorithm |
| title_full | LGNN: a novel linear graph neural network algorithm |
| title_fullStr | LGNN: a novel linear graph neural network algorithm |
| title_full_unstemmed | LGNN: a novel linear graph neural network algorithm |
| title_short | LGNN: a novel linear graph neural network algorithm |
| title_sort | lgnn a novel linear graph neural network algorithm |
| topic | graph neural network linear neural network graph deep learning graph representation learning high-order structural constraint |
| url | https://www.frontiersin.org/articles/10.3389/fncom.2023.1288842/full |
| work_keys_str_mv | AT shujuancao lgnnanovellineargraphneuralnetworkalgorithm AT shujuancao lgnnanovellineargraphneuralnetworkalgorithm AT shujuancao lgnnanovellineargraphneuralnetworkalgorithm AT shujuancao lgnnanovellineargraphneuralnetworkalgorithm AT xiaomingwang lgnnanovellineargraphneuralnetworkalgorithm AT zhonglinye lgnnanovellineargraphneuralnetworkalgorithm AT zhonglinye lgnnanovellineargraphneuralnetworkalgorithm AT zhonglinye lgnnanovellineargraphneuralnetworkalgorithm AT zhonglinye lgnnanovellineargraphneuralnetworkalgorithm AT mingyuanli lgnnanovellineargraphneuralnetworkalgorithm AT mingyuanli lgnnanovellineargraphneuralnetworkalgorithm AT mingyuanli lgnnanovellineargraphneuralnetworkalgorithm AT mingyuanli lgnnanovellineargraphneuralnetworkalgorithm AT haixingzhao lgnnanovellineargraphneuralnetworkalgorithm AT haixingzhao lgnnanovellineargraphneuralnetworkalgorithm AT haixingzhao lgnnanovellineargraphneuralnetworkalgorithm AT haixingzhao lgnnanovellineargraphneuralnetworkalgorithm |