HHSE: heterogeneous graph neural network via higher-order semantic enhancement

IF 3.3 3区 计算机科学 Q2 COMPUTER SCIENCE, THEORY & METHODS Computing Pub Date : 2024-01-22 DOI:10.1007/s00607-023-01246-x
Hui Du, Cuntao Ma, Depeng Lu, Jingrui Liu
{"title":"HHSE: heterogeneous graph neural network via higher-order semantic enhancement","authors":"Hui Du, Cuntao Ma, Depeng Lu, Jingrui Liu","doi":"10.1007/s00607-023-01246-x","DOIUrl":null,"url":null,"abstract":"<p>Heterogeneous graph representation learning has strong expressiveness when dealing with large-scale relational graph data, and its purpose is to effectively represent the semantic information and heterogeneous structure information of nodes in the graph. Current methods typically use shallow models to embed semantic information on low-order neighbor nodes in the graph, which prevents the complete retention of higher-order semantic feature information. To address this issue, this paper proposes a heterogeneous graph network for higher-order semantic enhancement called HHSE. Specifically, our model uses the identity mapping mechanism of residual attention at the node feature level to enhance the information representation of nodes in the hidden layer, and then utilizes two aggregation strategies to improve the retention of high-order semantic information. The semantic feature level aims to learn the semantic information of nodes in various meta path subgraphs. Extensive experiments on node classification and node clustering on three real-existing datasets show that the proposed approach makes practical improvements compared to the state-of-the-art methods. Besides, our method is applicable to large-scale heterogeneous graph representation learning.</p>","PeriodicalId":10718,"journal":{"name":"Computing","volume":"35 1 1","pages":""},"PeriodicalIF":3.3000,"publicationDate":"2024-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computing","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1007/s00607-023-01246-x","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, THEORY & METHODS","Score":null,"Total":0}
引用次数: 0

Abstract

Heterogeneous graph representation learning has strong expressiveness when dealing with large-scale relational graph data, and its purpose is to effectively represent the semantic information and heterogeneous structure information of nodes in the graph. Current methods typically use shallow models to embed semantic information on low-order neighbor nodes in the graph, which prevents the complete retention of higher-order semantic feature information. To address this issue, this paper proposes a heterogeneous graph network for higher-order semantic enhancement called HHSE. Specifically, our model uses the identity mapping mechanism of residual attention at the node feature level to enhance the information representation of nodes in the hidden layer, and then utilizes two aggregation strategies to improve the retention of high-order semantic information. The semantic feature level aims to learn the semantic information of nodes in various meta path subgraphs. Extensive experiments on node classification and node clustering on three real-existing datasets show that the proposed approach makes practical improvements compared to the state-of-the-art methods. Besides, our method is applicable to large-scale heterogeneous graph representation learning.

Abstract Image

查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
HHSE:通过高阶语义增强实现异构图神经网络
异构图表示学习在处理大规模关系图数据时具有很强的表现力,其目的是有效表示图中节点的语义信息和异构结构信息。目前的方法通常使用浅层模型将语义信息嵌入图中的低阶相邻节点,从而无法完整保留高阶语义特征信息。为解决这一问题,本文提出了一种用于高阶语义增强的异构图网络,称为 HHSE。具体来说,我们的模型在节点特征层利用剩余注意力的身份映射机制来增强隐层节点的信息表征,然后利用两种聚合策略来提高高阶语义信息的保留率。语义特征层旨在学习各种元路径子图中节点的语义信息。在三个真实数据集上进行的节点分类和节点聚类的广泛实验表明,与最先进的方法相比,我们提出的方法有了切实的改进。此外,我们的方法还适用于大规模异构图表示学习。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 去求助
来源期刊
Computing
Computing 工程技术-计算机:理论方法
CiteScore
8.20
自引率
2.70%
发文量
107
审稿时长
3 months
期刊介绍: Computing publishes original papers, short communications and surveys on all fields of computing. The contributions should be written in English and may be of theoretical or applied nature, the essential criteria are computational relevance and systematic foundation of results.
期刊最新文献
Mapping and just-in-time traffic congestion mitigation for emergency vehicles in smart cities Fog intelligence for energy efficient management in smart street lamps Contextual authentication of users and devices using machine learning Multi-objective service composition optimization problem in IoT for agriculture 4.0 Robust evaluation of GPU compute instances for HPC and AI in the cloud: a TOPSIS approach with sensitivity, bootstrapping, and non-parametric analysis
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1