基于多通道图卷积自编码器的图表示学习

doi:10.19678/j.issn.1000-3428.0063898

摘要/Abstract

摘要： 针对基于图卷积的自编码器模型对原始图属性和拓扑信息的保留能力有限、无法学习结构和属性之间深度关联信息等问题，提出基于多通道图卷积自编码器的图表示学习模型。设计拓扑和属性信息保留能力实验，验证了基于图卷积的自编码器模型具备保留节点属性和拓扑结构信息的能力。构建特定信息卷积编码器和一致信息卷积编码器，提取图的属性空间特征、拓扑空间特征以及两者关联特征，生成属性嵌入、拓扑嵌入和一致性嵌入，同时建立与编码器对称的卷积解码器，还原编码器过程。使用重构损失、局部约束和一致性约束，优化各编码器生成的低维嵌入表示。最终将蕴含不同图信息的多种嵌入进行融合，生成各节点的嵌入表示。实验结果表明，该模型在BlogCatalog和Flickr数据集上节点分类的Micro-F1和Macro-F1明显高于基线模型，在Citeseer数据集上节点聚类的精度和归一化互信息相比于表现最优的基线模型提升了11.84%和34.03%。上述实验结果证明了该模型采用的多通道方式能够在低维嵌入中保留更丰富的图信息，提升图机器学习任务的性能表现。

关键词: 图表示学习, 图卷积网络, 自编码器, 节点分类, 节点聚类

Abstract: This study proposes a graph representation learning model based on multi-channel graph convolutional autoencoders to address the limited ability of graph convolutional autoencoders in fusing node attributes and graph topology, and their inability to learn deep associations between node attributes.First, design topology and attribute information retention capability experiments are designed to verify the ability of a graph convolutional autoencoder in retaining node attribute and topological structure information.Second, specific and consensus convolutional encoders are designed to extract attribute- and topology-space features and their association, as well as to generate attribute, topology, and consensus embeddings.Third, convolutional decoders symmetric to the encoders are designed for recovering the encoder process.Fourth, reconstruction loss, local and consensus constraints are introduced to optimize low-dimensional embeddings generated by different encoders.Finally, multiple embeddings that contain different graph information are fused to generate an embedding representation for each node.The proposed model performs better than the baseline models in terms of node classification and node clustering.Its Micro-F1 and Macro-F1 for node classification are significantly higher than those of the baseline models of the BlogCatalog and Flickr datasets.Meanwhile, its Clustering Accuracy(Cluster-Acc) and Normalized Mutual Information(NMI) for node clustering on the Citeseer dataset are 11.84% and 34.03% higher, respectively, than the best-performing baseline model.The results show that the multi-channel approach adopted in the proposed model can retain richer graph information in low-dimensional embedding and improve downstream task performance.

Key words: graph representation learning, Graph Convolution Network(GCN), autoencoder, node classification, node clustering

中图分类号:

TP183

袁立宁, 胡皓, 刘钊. 基于多通道图卷积自编码器的图表示学习[J]. 计算机工程, 2023, 49(2): 150-160,174.

YUAN Lining, HU Hao, LIU Zhao. Graph Representation Learning Based on Multi-Channel Graph Convolutional Autoencoders[J]. Computer Engineering, 2023, 49(2): 150-160,174.

http://www.ecice06.com/CN/Y2023/V49/I2/150

图/表 11

20230216181035

20230216181039

20230216181042

20230216181046

20230216181050

20230216181054

20230216181057

20230216181103

20230216181106

20230216181110

20230216181113

参考文献

[1] PRAKASH S K A, TUCKER C S.Node classification using kernel propagation in graph neural networks[J].Expert Systems with Applications, 2021, 174:114655.
[2] 邱少明, 於涛, 杜秀丽, 等.基于节点多属性相似性聚类的社团划分算法[J].计算机工程, 2020, 46(7):84-90, 97. QIU S M, YU T, DU X L, et al.Community division algorithm based on similarity clustering of node multiple attribute[J].Computer Engineering, 2020, 46(7):84-90, 97.(in Chinese)
[3] HUANG Z H, WU J X, ZHU W T, et al.Visualizing complex networks by leveraging community structures[J].Physica A:Statistical Mechanics and Its Applications, 2021, 565:125506.
[4] BERAHMAND K, NASIRI E, ROSTAMI M, et al.A modified Deepwalk method for link prediction in attributed social network[J].Computing, 2021, 103(10):2227-2249.
[5] WANG Y L, CUI L Y, ZHANG Y.Improving skip-gram embeddings using BERT[J].IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2021, 29:1318-1328.
[6] YOU X M, et al.Representation method of cooperative social network features based on Node2Vec model[J].Computer Communications, 2021, 173:21-26.
[7] 马扬, 程光权, 梁星星, 等.有向加权网络中的改进SDNE算法[J].计算机科学, 2020, 47(4):233-237. MA Y, CHENG G Q, LIANG X X, et al.Improved SDNE in weighted directed network[J].Computer Science, 2020, 47(4):233-237.(in Chinese)
[8] LEVADA A L M, HADDAD M F C.Entropic Laplacian eigenmaps for unsupervised metric learning[C]//Proceedings of the 34th SIBGRAPI Conference on Graphics, Patterns and Images.Washington D.C., USA:IEEE Press, 2021:307-314.
[9] GAO H C, HUANG H.Deep attributed network embedding[C]//Proceedings of the 27th International Joint Conference on Artificial Intelligence.Stockholm, Sweden:International Joint Conferences on Artificial Intelligence Organization, 2018:3364-3370.
[10] WU Z H, PAN S R, CHEN F W, et al.A comprehensive survey on graph neural networks[J].IEEE Transactions on Neural Networks and Learning Systems, 2020, 32(1):4-24.
[11] CHEN Z M, WEI X S, WANG P, et al.Learning graph convolutional networks for multi-label recognition and applications[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 3496:1-8.
[12] YE Y, JI S H.Sparse graph attention networks[J].IEEE Transactions on Knowledge and Data Engineering, 2023, 35(1):905-916.
[13] GUO Z H, WANG F, YAO K X, et al.Multi-scale variational graph autoencoder for link prediction[C]//Proceedings of the 15th ACM International Conference on Web Search and Data Mining.New York, USA:ACM Press, 2022:334-342.
[14] ISLAM Z, ABDEL-ATY M, CAI Q, et al.Crash data augmentation using variational autoencoder[J].Accident Analysis & Prevention, 2021, 151:105950.
[15] PARK J, LEE M, CHANG H J, et al.Symmetric graph convolutional autoencoder for unsupervised graph representation learning[C]//Proceedings of IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2019:6518-6527.
[16] LI Q M, HAN Z C, WU X M.Deeper insights into graph convolutional networks for semi-supervised learning[C]//Proceedings of AAAI Conference on Artificial Intelligence.Palo Alto, USA:AAAI Press, 2018:3538-3545.
[17] WANG X, ZHU M Q, BO D Y, et al.AM-GCN:adaptive multi-channel graph convolutional networks[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining.New York, USA:ACM Press, 2020:1243-1253.
[18] LIU X Y, YANG B, CHEN H C, et al.A scalable redefined stochastic blockmodel[J].ACM Transactions on Knowledge Discovery from Data, 2021, 15(3):28-46.
[19] PEI H, WEI B, CHANG K C-C, et al.Geom-GCN:geometric graph convolutional networks[EB/OL].[2022-02-08].https://arxiv.org/abs/2002.05287v1.
[20] HU F Y.GraphAIR:graph representation learning with neighborhood aggregation and interaction[J].Pattern Recognition, 2021, 112:107745.
[21] CHEN H C, PEROZZI B, HU Y F, et al.HARP:hierarchical representation learning for networks[C]//Proceedings of AAAI Conference on Artificial Intelligence.Palo Alto, USA:AAAI Press, 2018:2127-2134.
[22] HUANG X, LI J D, HU X.Label informed attributed network embedding[C]//Proceedings of the 10th ACM International Conference on Web Search and Data Mining.New York, USA:ACM Press, 2017:731-739.
[23] 陈世聪, 袁得嵛, 黄淑华, 等.基于结构深度网络嵌入模型的节点标签分类算法[J].计算机科学, 2022, 49(3):105-112. CHEN S C, YUAN D Y, HUANG S H, et al.Node label classification algorithm based on structural depth network embedding model[J].Computer Science, 2022, 49(3):105-112.(in Chinese)
[24] XU X L, XU H Y, WANG Y, et al.AENEA:a novel autoencoder-based network embedding algorithm[J].Peer-to-Peer Networking and Applications, 2021, 14(3):1829-1840.
[25] SALEHI A, DAVULCU H.Graph attention auto-encoders[C]//Proceedings of the 32nd International Conference on Tools with Artificial Intelligence.Washington D.C., USA:IEEE Press, 2019:989-996.
[26] WANG J, LIANG J Y, YAO K X, et al.Graph convolutional autoencoders with co-learning of graph structure and node attributes[J].Pattern Recognition, 2022, 121:108215.
[27] 袁立宁, 刘钊.基于One-Shot聚合自编码器的图表示学习[J/OL].计算机应用:1-8[2021-12-31].http://kns.cnki.net/kcms/detail/51.1307.TP.20220105.1002.002.html. YUAN L N, LIU Z.Graph representation learning using autoencoder with One-Shot aggregation[J/OL].Computer Application:1-8[2022-02-08].http://kns.cnki.net/kcms/detail/51.1307.TP.20220105.1002.002.html (in Chinese).
[28] LEE Y, HWANG J W, LEE S, et al.An energy and GPU-computation efficient backbone network for real-time object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops.Washington D.C., USA:IEEE Press, 2019:752-760.
[29] SHOPOV V K, MARKOVA V D.Application of Hungarian algorithm for assignment problem[C]//Proceedings of International Conference on Information Technologies.Washington D.C., USA:IEEE Press, 2021:1-4.
[30] YANG F, LIU S F, DOBRIBAN E, et al.How to reduce dimension with PCA and random projections?[J].IEEE Transactions on Information Theory, 2021, 67(12):8154-8189.
[31] DAOUDI S, ZOUAOUI C M A, EL-MEZOUAR M C, et al.Parallelization of the K-means++ clustering algorithm[J].Ingénierie Des Systèmes d Information, 2021, 26(1):59-66.

选择文件类型/文献管理软件名称

选择包含的内容