基于因果推断的图注意力网络

doi:10.11896/jsjkx.220600230

计算机科学 ›› 2023, Vol. 50 ›› Issue (6A): 220600230-9.doi: 10.11896/jsjkx.220600230

基于因果推断的图注意力网络

张涛, 程毅飞, 孙欣煦

燕山大学信息科学与工程学院河北秦皇岛 066004;
河北省信息传输与信号处理重点实验室河北秦皇岛 066004

出版日期:2023-06-10 发布日期:2023-06-12
通讯作者: 张涛(zhtao@ysu.edu.cn)
基金资助:
国家自然科学基金(62176229);河北省自然科学基金(F2020203010)

Graph Attention Networks Based on Causal Inference

ZHANG Tao, CHENG Yifei, SUN Xinxu

School of Information Science and Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China;
Hebei Key Laboratory of Information Transmission and Signal Processing,Qinhuangdao,Hebei 066004,China

Online:2023-06-10 Published:2023-06-12
About author:ZHANG Tao,born in 1979,Ph.D,professor,doctoral supervisor,is a member of China Computer Federation.His main research interests include the causal inference,machine learning,and formal concept analysis.
Supported by:
National Natural Science Foundation of China(62176229) and Natural Science Foundation of Hebei Province,China(F2020203010).

摘要/Abstract

摘要： 图注意力网络(Graph Attention Network,GAT)是一种重要的图神经网络,在分类任务中有着广泛的应用。但是当图中邻域节点受到干扰时,模型分类准确度会受影响而降低。对此,提出一种基于因果推断的图注意力网络(Causal Graph Attention Network,C-GAT)以提升网络的鲁棒性。该模型首先计算目标节点的邻域与其标签之间的因果权重,并以此对邻域进行因果采样;然后计算采样后邻域与目标节点之间的注意力系数;最后根据注意力系数对邻域信息进行加权聚合,获得目标节点的嵌入特征。在Cora,Pubmed和Citeseer数据集上的实验结果表明,在无扰动的情况下,C-GAT的分类性能与经典模型持平;在有扰动的情况下,相比于经典模型,C-GAT的分类准确度至少提升了6%,在鲁棒性和时间成本上有着较好的平衡。

关键词: 图注意力网络, 因果推断, 注意力机制, 因果权重

Abstract: Graph attention network(GAT) is an important graph neural network with a wide range of applications in classification tasks.However,when the neighborhood nodes in the graph are disturbed,the model classification accuracy will be affected and degraded.In response,a graph attention network based on causal inference named causal graph attention network(C-GAT) is proposed to improve the robustness of the network.The model first calculates the causal weights between the neighborhood of the target node and its label and uses them to sample the neighborhood.Then the attention coefficient between the sampled neighborhood and the target node is calculated.Finally,the embedding features of the target nodes are obtained by weighted aggregation of the neighborhood information based on the attention coefficients.Experimental results on the Cora,Pubmed and Citeseer datasets show that the classification performance of C-GAT is on par with the classical model in the case of no perturbation.In the presence of perturbations,the classification accuracy of C-GAT improves by at least 6% compared to the classical model,with a better balance of robustness and time cost.

Key words: Graph attention networks, Causal inference, Attention mechanism, Causal weight

中图分类号:

TP183

张涛, 程毅飞, 孙欣煦. 基于因果推断的图注意力网络[J]. 计算机科学, 2023, 50(6A): 220600230-9. https://doi.org/10.11896/jsjkx.220600230

ZHANG Tao, CHENG Yifei, SUN Xinxu. Graph Attention Networks Based on Causal Inference[J]. Computer Science, 2023, 50(6A): 220600230-9. https://doi.org/10.11896/jsjkx.220600230

参考文献

[1]VELIKOVI P,CUCURULL G,CASANOVA A,et al.GraphAttention Networks[C]//The 6th International Conference on Learning Representations.2018:1-12.
[2]TAN Y,WANG J,ZHANG C.Review of Text ClassificationMethods Based on Graph Convolutional Network[J].Computer Science,2022,49(8):205-216.
[3]XU S J,LIU Q Y,SHI Y,et al.Person Re-Identification Based on Diversified Local Attention Network[J].Journal of Electroni-cs & Information Technology,2022,44(1):211-220.
[4]JUNG J,HEO H S,YU H J,et al.Graph Attention Networks for Speaker Verification[C]//IEEE International Conference on Acoustics,Speech and Signal Processing.Ontario:IEEE,2021:6149-6153.
[5]HAN H,WU Y H,QIN X Y.An Interactive Graph AttentionNetworks Model for Aspect-level Sentiment Analysis[J].Journal of Electronics & Information Technology,2021,43(11):3282-3290.
[6]ZHANG J,LI M X,GAO K S,et al.Word and Graph Attention Networks for Semi-Supervised Classification[J].Knowledge and Information Systems,2021,63:2841-2859.
[7]CAI W W,WEI Z G.Remote Sensing Image Classification Based on a Cross-Attention Mechanism and Graph Convolution[J].IEEE Geoscience and Remote Sensing Letters,2022,19:1-5.
[8]ZHOU H,YANG Y Z,LUO T J,et al.A Unified Deep Sparse Graph Attention Network for Scene Graph Generation[J].Pattern Recognition,2022,123:108367.
[9]YANG Y D,WANG X C,SONG M L,et al.SPAGAN:Shortest Path Graph Attention Network[C]//International Joint Confe-rence on Artificial Intelligence.2019:4099-4105.
[10]ZHOU A Z,LI Y F.Structural Attention Network for Graph[J].Applied Intelligence,2021,51(8):6255-6264.
[11]BAI J,DING B X,XIAO Z,et al,Hyperspectral Image Classification Based on Deep Attention Graph Convolutional Network[J].IEEE Transactions on Geoscience and Remote Sensing,2022,60:1-16.
[12]GU W W,GAO F,LOU X D,et al.Discovering Latent Node Information by Graph Attention Network[J].Scientific Reports,2021,11(1):6979.
[13]ZHANG K,ZHU Y,WANG J,et al.Adaptive Structural Fin-gerprints for Graph Attention Networks[C]//The 8th International Conference on Learning Representations.2020.
[14]YANG L,LI W X,GUO Y F,et al.Graph-CAT:Graph Co-Attention Networks Via Local and Global Attribute Augmentations[J].Future Generation Computer Systems,2021,118:170-179.
[15]XIE Y,ZHANG Y Q,GONG M G,et al.Multi-View Graph Attention Networks[J].Neural Networks,2020,132:180-189.
[16]ZHANG Z C,LI M,LIN X,et al.Multistep Speed Prediction on Traffic Networks:A Deep Learning Approach ConsideringSpatio-Temporal Dependencies[J].Transportation Research Part C:Emerging Technologies,2019,105:297-322.
[17]JI C J,WANG R X,ZHU R X,et al.Hop-Aware SupervisionGraph Attention Networks for Sparsely Labeled Graphs[J/OL].https://arxiv.org/abs/2004.04333.
[18]ZHANG H M,XU M.Graph Neural Networks with MultipleKernel Ensemble Attention[J].Knowledge-Based Systems,2021,229:107299.
[19]BAI S,ZHANG F H,TORR P H S.Hypergraph Convolution and Hypergraph Attention[J].Pattern Recognition,2021,110:107637.
[20]WANG Z H,SHEN H W,CAO Q,et al.Survey on Graph Classification[J].Journal of Software,2022,33(1):171-192.
[21]FENG W Z,ZHANG J,DONG Y X,et al.Graph Random Neural Networks for Semi-supervised Learning on Graphs[C]//Advances in Neural Information Processing Systems,2020,33:22092-22103.
[22]JUDEA P.The Seven Tools of Causal Inference with Reflections on Machine Learning[J].Communications of the ACM,2019,62(3):54-60.
[23]JUDEA P.Causal Inference[M].Causality:Objectives and Assessment,2010:39-58.
[24]SCHOLKOPF B,LOCATELLO F,BAUER S,et al.TowardCausal Representation Learning[J].Proceedings of the IEEE,2021,109(5):612-634.
[25]RICHENS J G,LEE C M,JOHRI S.Improving the Accuracy of Medical Diagnosis with Causal Machine Learning[J].Nature Communications,2020,11:47-54.
[26]LITTLE M A,BADAWY R.Causal Bootstrapping[J/OL].https://arxiv.org/abs/1910.09648.
[27]LIU L,WANG S,HU B,et al.Learning structures of interval-based Bayesian networks in probabilistic generative model for human complex activity recognition[J].Pattern Recognition,2018,81:545-561
[28]ZHANG T,LIU M Q,LIU W Y.The Causality Research Between Syndrome Elements by Attribute Topology[J].Computational and Mathematical Methods in Medicine,2018,2018:1-12.
[29]KINGMA D P,BA J L.Adam:A Method for Stochastic Optimization[C]//ICIR 2015.2015.
[30]KIPF T N,WELLING M.Semi-supervised Classification withGraph Convolutional Networks[C]//International Conference on Learning Representations.2017.
[31]HAMILTON W L,YING R,LESKOVEC J.Inductive Representation Learning on Large Graphs[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.2017:1025-1035.
[32]XIE T,WANG B,JAY K C C.GraphHop:An Enhanced Label Propagation Method for Node Classification[J/OL].https://arxiv.org/abs/2101.02326v1.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

基于因果推断的图注意力网络

Graph Attention Networks Based on Causal Inference

PDF (PC)

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

Metrics

本文评价

推荐阅读 0