《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (2): 357-364.DOI: 10.11772/j.issn.1001-9081.2021030380

• 人工智能 • 上一篇    

融合全局结构信息的拓扑优化图卷积网络

富坤(), 高金辉, 赵晓梦, 李佳宁   

  1. 河北工业大学 人工智能与数据科学学院,天津 300401
  • 收稿日期:2021-03-14 修回日期:2021-06-01 接受日期:2021-06-03 发布日期:2022-02-21 出版日期:2022-02-10
  • 通讯作者: 富坤
  • 作者简介:富坤(1979—),女,辽宁辽阳人,副教授,博士,主要研究方向:社会网络分析、网络表示学习;
    高金辉(1995—),男,天津人,硕士研究生,主要研究方向:社会网络分析、网络表示学习;
    赵晓梦(1995—),女,河南郑州人,硕士研究生,主要研究方向:网络表示学习;
    李佳宁(1996—),男,河北唐山人,硕士研究生,主要研究方向:网络表示学习。
  • 基金资助:
    国家自然科学基金资助项目(61806072)

Topology optimization based graph convolutional network combining with global structural information

Kun FU(), Jinhui GAO, Xiaomeng ZHAO, Jianing LI   

  1. School of Artificial Intelligence,Hebei University of Technology,Tianjin 300401,China
  • Received:2021-03-14 Revised:2021-06-01 Accepted:2021-06-03 Online:2022-02-21 Published:2022-02-10
  • Contact: Kun FU
  • About author:FU Kun, born in 1979, Ph. D., associate professor. Her research interests include social network analysis, network representation learning.
    GAO Jinhui, born in 1995, M. S. candidate. His research interests include social network analysis, network representation learning.
    ZHAO Xiaomeng, born in 1995, M. S. candidate. Her research interests include network representation learning.
    LI Jianing, born in 1996, M. S. candidate. His research interests include network representation learning.
  • Supported by:
    National Natural Science Foundation of China(61806072)

摘要:

基于拓扑优化的图卷积网络(TOGCN)是一类图卷积神经网络(GCNN)模型,它通过网络中的辅助信息优化网络拓扑结构,有利于反映节点间的联系程度;然而TOGCN模型仅注重局部节点之间的关联关系,对网络潜在的全局结构信息关注不足。融合全局特征信息,有助于提高模型的性能和处理信息缺失时的鲁棒性。提出了融合全局结构信息的拓扑优化图卷积网络(GE-TOGCN)模型,该模型一方面利用相邻节点的属性对拓扑图进行优化;另一方面使用类信息作为网络的全局结构信息,从而保持类内聚合性和类间分离性。首先根据标记节点计算类中心向量;然后利用部分未标记节点来更新类中心向量;最后将所有节点根据其与类中心向量的相似度分配到对应的类中,并通过一个半监督损失函数优化各类的类中心向量与节点的最终表示向量。在Cora、Citeseer数据集上,在标签信息缺失的情况下运用得到的节点表示向量进行了节点分类任务与节点可视化任务。实验结果表明,GE-TOGCN模型与图卷积网络(GCN)、图学习卷积网络(GLCN)等模型相比,在Cora数据集上的分类准确率提高了1.2~12.0个百分点,在Citeseer数据集上的分类准确率提高了0.9~9.9个百分点;而在节点可视化任务中所提模型的类内节点聚合程度更高,类簇之间的边界更明显。可见,融合类全局信息能减少标签信息缺失对模型学习效果的不良影响,且该模型得到的节点表示在下游任务中表现出了更好的性能。

关键词: 网络表示学习, 图嵌入, 图卷积神经网络, 全局结构信息, 拓扑优化

Abstract:

As a kind of Graph Convolutional Neural Network (GCNN), Topology Optimization based Graph Convolutional Network (TOGCN) model adopts auxiliary information in the network to optimize topological structure of the network, thereby helping to reflect the relational degrees between the nodes. However, TOGCN model only focuses on the association between local nodes, and not enough on the potential global structure information. Fusing global feature information, the model will help to improve performance as well as its robustness in dealing with incomplete information. A Global structure information Enhanced-TOGCN (GE-TOGCN) model was proposed, the attributes of neighboring nodes were utilized to optimize the topological graph, and the class information was regarded as the global structure information to maintain intra-class aggregation and inter-class separation. Firstly, the center vector of each class was calculated by the labeled nodes, then some unlabeled nodes were selected to update these class center vectors. Finally, all the nodes were assigned to the corresponding class according to their similarity to class center vectors, and a semi-supervised loss function was adopted to optimize the class center vector of each class and the final representation vectors of the nodes. On Cora and Citeseer datasets, node classification task and node visualization task were performed by using the obtained node representation vectors with the loss of label information. Experimental results show that compared with Graph Convolutional Network (GCN), Graph Learning-Convolutional Network (GLCN) and other models, GE-TOGCN has the classification accuracy increased by 1.2-12.0 percentage points on Cora dataset, and the classification accuracy increased by 0.9-9.9 percentage points on Citeseer dataset. In node visualization task, the proposed model has higher degree of intra-class node aggregation and more obvious boundaries between class clusters. In summary, the fusion of class global information can reduce the negative influence of label information loss on learning effects of the model, and the node representations obtained by the proposed model have better performance in downstream tasks.

Key words: network representation learning, graph embedding, Graph Convolutional Neural Network (GCNN), global structural information, topology optimization

中图分类号: