Heterogeneous information network embedding method integrating incomplete multi-view

Heterogeneous Information Network (HIN) embedding maps complex heterogeneous information to a low-dimensional dense vector space, which is conducive to the calculation and storage of network data. The existing multi-view-based HIN embedding method considers multiple semantic relationships between no...

Full description

Saved in:
Bibliographic Details
Published inJi suan ji ke xue Vol. 48; no. 9; p. 68
Main Authors Zheng, Susu, Kanto, Sea, Yuan, Weiwei
Format Journal Article
LanguageChinese
Published Chongqing Guojia Kexue Jishu Bu 01.01.2021
Subjects
Correlation analysis
Embedding
Missing data
Nodes
Semantics
Online AccessGet full text

Cover

More Information
Summary:Heterogeneous Information Network (HIN) embedding maps complex heterogeneous information to a low-dimensional dense vector space, which is conducive to the calculation and storage of network data. The existing multi-view-based HIN embedding method considers multiple semantic relationships between nodes, but ignores the incompleteness of the view. Most views have missing data, and the direct fusion of multiple incomplete views will lead to poor embedding effects. For this reason, this paper proposes a HIN embedding method (Incomplete Multi-view Fusion Based HIN Embedding, IMHE). The key idea of ​​IMHE is to aggregate neighbors of other views to reconstruct an incomplete view. Since different single views describe the same network, neighbors in other views can restore the structural information of incomplete views to a certain extent. IMHE first generates a sequence of nodes in different views, and uses the multi-head attention method to learn single-view embedding. For each incomplete view, IMHE finds the k-or
ISSN:1002-137X