GraPASA: Parametric Graph Embedding via Siamese Architecture

Embargo End Date
2021-10-25

Type
Article

Authors
Chen, Yujun
Sun, Ke
Pu, Juhua
Xiong, Zhang
Zhang, Xiangliang

KAUST Department
Computer Science Program
Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division

KAUST Grant Number
Award No. 2639

Online Publication Date
2019-10-25

Print Publication Date
2019-10

Date
2019-10-25

Abstract
Graph representation learning or graph embedding is a classical topic in data mining. Current embedding methods are mostly non-parametric, where all the embedding points are unconstrained free points in the target space. These approaches suffer from limited scalability and an over-flexible representation. In this paper, we propose a parametric graph embedding by fusing graph topology information and node content information. The embedding points are obtained through a highly flexible non-linear transformation from node content features to the target space. This transformation is learned using the contrastive loss function of the siamese network to preserve node adjacency in the input graph. On several benchmark network datasets, the proposed GraPASA method shows a significant margin over state-of-the-art techniques on benchmark graph representation tasks.

Citation
Chen, Y., Sun, K., Pu, J., Xiong, Z., & Zhang, X. (2019). GraPASA: Parametric Graph Embedding via Siamese Architecture. Information Sciences. doi:10.1016/j.ins.2019.10.027

Acknowledgements
This work was partially supported and funded by King Abdullah University of Science and Technology (KAUST) through the KAUST Office of Sponsored Research (OSR) under Award No. 2639, National Key R&D Program of China (2017YFC0803700), National Natural Science Foundation of China (61502320), Science Foundation of Shenzhen City in China (JCYJ20160419152942010), the State Key Laboratory of Software Development Environment, and Aeronautical Science Foundation of China.

Publisher
Elsevier BV

Journal
Information Sciences

DOI
10.1016/j.ins.2019.10.027

Additional Links
https://linkinghub.elsevier.com/retrieve/pii/S0020025519309806

Permanent link to this record