ABSTRACT
In order to accurately reflect the information of each node's neighbours and the topological relationship of the graph structure, the graph neural network node classification task aims to treat each node and all of its neighbouring nodes as a subgraph and update the feature representation of each node through information transfer. However, a crucial aspect of network symmetry is disregarded in the current research of the graph neural network node classification job, resulting in redundant data in the training dataset. Consequently, this research suggests a strategy for classifying nodes in a graph neural network based on graph degree-symmetry (The same abbreviation later is GDS). The next two steps make up the majority of the method: Using the nauty technique to retrieve the network's track information, calculating the similarity between nodes according to the degree-symmetry, then applying the softmax function for normalization, assigning weights to each neighboring node; Lastly, neighborhood node information is combined, the node information is updated using weighted summation to produce graph data with more useful node characteristics. The revised network is then re-entered into the graph neural network for convolution, which increases the accuracy of the classification task. In this paper, GDS is equipped with SIGN, MixHop, SGC models to improve the performance. Three publicly accessible citation network datasets were used for the experimental analysis of GDS fitted with the three graph neural network models mentioned above. The findings of the analysis confirmed the efficacy of GDS, showing that GDS can more effectively utilize the semantic information of graphs to synthesize more reasonable node feature representations to improve node classification.
- Zhang D, Jie Y, Zhu X, Network Representation Learning: A Survey[J]. IEEE Transactions on Big Data, 2017, PP (99):1-1Google Scholar
- B.B. Xu, C.T. Sham, J.J. Huang, A review of graph convolutional neural networks[J]. Journal of Computer Science, 2020, 43(5):26Google Scholar
- Frasca F, Rossi E, Eynard D, SIGN: Scalable Inception Graph Neural Networks[J]. 2020.DOI:10.48550/arXiv.2004.11198Google ScholarCross Ref
- Abu-El-Haija S, Perozzi B, Kapoor A, MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing: PMLR 2019Google Scholar
- Wu F, Zhang T, Souza A H D, Simplifying Graph Convolutional Networks[J]. 2019.DOI:10.48550/arXiv.1902.07153Google ScholarCross Ref
- Velikovi P, Cucurull G, Casanova A, Graph Attention Networks[J]. 2017Google Scholar
- Kipf T N, Welling M. Semi-Supervised Classification with Graph Convolutional Networks[J]. 2016Google Scholar
- Hamilton W L, Ying R, Leskovec J. Inductive Representation Learning on Large Graphs[J]. 2017Google Scholar
- Weston, J., Ratle, F., Mobahi, H., and Collobert, R. Deeplearning via semi-supervised embedding. In Neural Networks: Tricks of the Trade, pp. 639–655, 2012Google ScholarCross Ref
- Hassani K, Khasahmadi A H. Contrastive Multi-View Representation Learning on Graphs[J]. 2020Google Scholar
- Feng W, Dong Y, Huang T, GRAND+: Scalable Graph Random Neural Networks[J]. 2022Google Scholar
- Rong Y, Huang W, Xu T, DropEdge: Towards Deep Graph Convolutional Networks on Node Classification[J]. 2019Google Scholar
- Zou X, Jia Q, Zhang J, Dimensional Reweighting Graph Convolutional Networks[J]. 2019Google Scholar
- Chen M, Wei Z, Huang Z, Simple and Deep Graph Convolutional Networks[J]. 2020Google Scholar
- Perozzi B, Al-Rfou R, Skiena S. DeepWalk: Online Learning of Social Representations[J].ACM, 2014.DOI:10.1145/2623330.2623732Google ScholarDigital Library
- Tang J, Qu M, Wang M, LINE: Large-scale Information Network Embedding.2015[2023-07-05].DOI:10.1145/2736277.2741093Google ScholarDigital Library
- Grover A, Leskovec J. node2vec: Scalable Feature Learning for Networks[J].ACM, 2016.DOI:10.1145/2939672.2939754Google ScholarDigital Library
- Wu, W.-T. Application of graph symmetry theory to some important problems in social network analysis[D]. Fudan University,2010Google Scholar
- Wikipedia F. Group theory[J]. Group Theory, 2011Google Scholar
- G Sardanashvily. Gauge group (mathematics). 2013Google Scholar
- B. D. McKay, Practical graph isomorphism, Congr. Numer. 30 (1981), 45–87Google Scholar
- Liao Zifeng. Network symmetry and its application to protein interaction networks[D]. Xi'an University of Electronic Science and Technology, 2012Google Scholar
Index Terms
- Node Classification of Graph Neural Networks Based on Graph Degree-Symmetry
Recommendations
Label-Consistency based Graph Neural Networks for Semi-supervised Node Classification
SIGIR '20: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information RetrievalGraph neural networks (GNNs) achieve remarkable success in graph-based semi-supervised node classification, leveraging the information from neighboring nodes to improve the representation learning of target node. The success of GNNs at node ...
Graph Neural Networks with Information Anchors for Node Representation Learning
AbstractIn the era of big data, the large-scale information network applications need to process and analyze increasingly complex graph structure relationships. However, traditional methods of representing network structures are difficult to reflect ...
DEMO-Net: Degree-specific Graph Neural Networks for Node and Graph Classification
KDD '19: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data MiningGraph data widely exist in many high-impact applications. Inspired by the success of deep learning in grid-structured data, graph neural network models have been proposed to learn powerful node-level or graph-level representation. However, most of the ...
Comments