基于嵌套生成对抗学习的网络嵌入

doi:10.12263/DZXB.20210592

摘要/Abstract

摘要：

当前网络嵌入研究更多关注信息网络结构和结点之间一阶或高阶近似关系，对于网络结点自身属性考虑较少.本文提出一种嵌套的生成对抗网络模型N-GAN（Nesting Generative Adversarial Networks for Network Embedding），实现了网络结构和节点属性同时嵌入到低维向量，从而最大程度保存原始高维信息网络特征.N-GAN模型设计灵活，具有很好的延伸性和扩张性，并在真实数据上验证了N-GAN的性能及其稳定性，其嵌入的低维表示在不同应用中表现出不错的性能.

关键词: 数据挖掘, 网络嵌入, 生成对抗学习, 信息网络

Abstract:

The current network embedding researches focus more on the information network structure and first-order or higher-order approximation of nodes, but less on the attributes of network nodes. This paper proposes a nested generative adversarial network model N-GAN(Nesting Generative Adversarial Networks for Network Embedding), which embeds the network structure and nodes' attributes into the low-dimensional vector at the same time, so as to preserve the feature of the original high-dimensional information network maximumly. N-GAN model is flexible in design and has good extensibility and expansibility. The performance and stability of N-GAN model are verified on real datasets. The embedded low-dimensional representation of N-GAN model shows good performance in different tasks.

Key words: data mining, network embedding, generative adversarial learning, information network

中图分类号:

TP391

沈鹏飞, 徐臻, 王英. 基于嵌套生成对抗学习的网络嵌入[J]. 电子学报, 2022, 50(9): 2155-2163.

SHEN Peng-fei, XU Zhen, WANG Ying. Network Embedding Based on Nested Generative Adversarial Networks[J]. Acta Electronica Sinica, 2022, 50(9): 2155-2163.

图/表 9

参考文献 28

1	TANG J. Computational models for social network analysis: A brief survey[C]//Proceedings of the 26th International Conference on World Wide Web Companion- WWW'17 Companion. New York: ACM Press, 2017: 921-925.
2	CEN Y K, ZOU X, ZHANG J W, et al. Representation learning for attributed multiplex heterogeneous network[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York: ACM, 2019: 1358-1368.
3	SUN G L. New progress in research and application of machine learning[J]. Chinese Journal of Electronics, 2020, 29(6): 991-991.
4	SUN X, SONG Z H, DONG J Y, et al. Network structure and transfer behaviors embedding via deep prediction model[C]//Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2019: 5041-5048.
5	刘露, 胡封晔, 牛亮, 等. 异质网络中基于节点影响力的相似度度量方法[J]. 电子学报, 2019, 47(9): 1929-1936.
	LIU L, HU F Y, NIU L, et al. Node influence based similarity measure method in heterogeneous network[J]. Acta Electronica Sinica, 2019, 47(9): 1929-1936. (in Chinese)
6	WANG Z, YE X, WANG C, et al. RSDNE: Exploring relaxed similarity and dissimilarity from completely-imbalanced labels for network embedding[C]//Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2018: 475-482.
7	LAI Y Y, NEVILLE J, GOLDWASSER D. TransConv: relationship embedding in social networks[C]//Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2019: 4130-4138.
8	国琳, 左万利. 基于兴趣图谱的用户兴趣分布分析及专家发现[J]. 电子学报, 2015, 43(8): 1561-1567.
	GUO L, ZUO W L. Analysis of user interest distribution and expert finding based on interest graphs[J]. Acta Electronica Sinica, 2015, 43(8): 1561-1567. (in Chinese)
9	ASSUNÇÃO F, LOURENÇO N, MACHADO P, et al. DENSER: deep evolutionary network structured representation[J].Genetic Programming and Evolvable Machines, 2019, 20(1): 5-35.
10	ZHANG P Z, GONG M G, ZHANG H, et al. DRLnet: deep difference representation learning network and an unsupervised optimization framework[C]//Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2017: 3413-3419.
11	ZHANG D K, YIN J, ZHU X Q, et al. Homophily, structure, and content augmented network representation learning[C]//2016 IEEE 16th International Conference on Data Mining. Piscataway: IEEE, 2016: 609-618.
12	CAO S, LU W, XU Q. Deep neural networks for learning graph representations[C]//Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2016: 1145-1152.
13	张昱, 刘开峰, 张全新, 等. 基于组合-卷积神经网络的中文新闻文本分类[J]. 电子学报, 2021, 49(6): 1059-1067.
	ZHANG Y, LIU K F, ZHANG Q X, et al. A combined-convolutional neural network for Chinese news text classification[J]. Acta Electronica Sinica, 2021, 49(6): 1059-1067. (in Chinese)
14	TANG J L, GAO H J, HU X, et al. Exploiting homophily effect for trust prediction[C]//Proceedings of the sixth ACM International Conference on Web Search And Data Mining-WSDM'13. New York: ACM Press, 2013: 53-62.
15	CHENG K W, LI J D, LIU H. Unsupervised feature selection in signed social networks[C]//Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2017: 777-786.
16	WANG X, Cui P, WANG J, et al. Community preserving network embedding [C]//Proceedings of the 31st AAAI Conference on Artificial Intelligence. San Francisco, Palo Alto, CA: AAAI Press,2017: 203-209.
17	QIU J Z, DONG Y X, MA H, et al. Network embedding as matrix factorization: Unifying DeepWalk, LINE, PTE, and node2vec[C]//Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining. New York: ACM, 2018: 459-467.
18	PEROZZI B, AL-RFOU R, SKIENA S. DeepWalk: online learning of social representations[C]//Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2014: 701-710.
19	TANG J, QU M, WANG M Z, et al. LINE: large-scale information network embedding[EB/OL].(2015-05-12). .
20	TANG J, QU M, MEI Q Z. PTE: predictive text embedding through large-scale heterogeneous text networks[C]//Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2015: 1165-1174.
21	GROVER A, LESKOVEC J. node2vec: Scalable feature learning for networks[C]//Proceedings. International Conference on Knowledge Discovery & Data Mining. New York: ACM, 2016: 855-864.
22	Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems-Volume 2. Cambridge, USA: MIT Press, 2014:2672-2680.
23	MIRZA M, OSINDERO S. Conditional generative adversarial nets[EB/OL].(2014-11-06). .
24	Tolstikhin I, Gelly S, Bousquet O. AdaGAN: Boosting generative models[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook,NY:Curran Associates Inc, 2017: 5430-5439.
25	ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein Gan[EB/OL]. (2017-01-26). .
26	MAO X D, LI Q, XIE H R, et al. Least Squares generative adversarial networks[C]//2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 2813-2821.
27	权宇, 李志欣, 张灿龙, 等. 融合深度扩张网络和轻量化网络的目标检测模型[J]. 电子学报, 2020, 48(2): 390-397.
	QUAN Y, LI Z X, ZHANG C L, et al. Fusing deep dilated convolutions network and light-weight network for object detection[J]. Acta Electronica Sinica, 2020, 48(2): 390-397. (in Chinese)
28	RIBEIRO L F R, SAVERESE P H P, FIGUEIREDO D R. struc2vec: learning node representations from structural identity[C]//Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2017: 385-394.

数据集	arXiV-AstroPh	arXiv-GrQc
结点个数	18772	5242
链路个数	198110	14496
结点最大度数	504	81
平均聚类系数	0.6306	0.5296
网络密度	0.0011	0.0010

数据集	arXiV-AstroPh	arXiv-GrQc
结点个数	18772	5242
链路个数	198110	14496
结点最大度数	504	81
平均聚类系数	0.6306	0.5296
网络密度	0.0011	0.0010

数据集	Cora
结点个数	2708
链路个数	5278
结点类别数	7
结点最大度数	168
同一类别结点最大个数	818
同一类别结点最小个数	180
网络密度	0.0014

数据集	Cora
结点个数	2708
链路个数	5278
结点类别数	7
结点最大度数	168
同一类别结点最大个数	818
同一类别结点最小个数	180
网络密度	0.0014

Model	arXiV-AstroPh		arXiv-GrQc
Model	Acc	Macro-F1	Acc	Macro-F1
Deepwalk	0.9579	0.9306	0.9389	0.9149
LINE-O1	0.9250	0.8980	0.9120	0.8932
LINE-O2	0.9143	0.8712	0.9008	0.8655
struc2vec	0.9184	0.8393	0.9140	0.8352
N-GAN	0.9411	0.9158	0.9251	0.9031
GAN1	0.9307	0.9037	0.9137	0.8902
GAN2_a	0.9389	0.9134	0.9191	0.8970
GAN2_b	0.9382	0.9122	0.9202	0.8988