A comparative study on network alignment techniques

作者：

Highlights：

• Overcoming structural and attribute noises is challenging to the community.

• PALE and IONE are less sensitive to structural noise than spectral methods.

• REGAL is more resistant to attribute noise and have faster computation time.

• The size imbalance between source and target networks affects the alignment.

• Graph connectivity and connected components slightly affect alignment algorithms.

摘要

•Overcoming structural and attribute noises is challenging to the community.•PALE and IONE are less sensitive to structural noise than spectral methods.•REGAL is more resistant to attribute noise and have faster computation time.•The size imbalance between source and target networks affects the alignment.•Graph connectivity and connected components slightly affect alignment algorithms.

论文关键词：Network alignment,Graph matching,Network embedding,Graph mining,Node representation learning,Low-rank matrix factorization

论文评审过程：Received 21 April 2019, Revised 17 August 2019, Accepted 17 August 2019, Available online 28 August 2019, Version of Record 7 September 2019.

论文官网地址：https://doi.org/10.1016/j.eswa.2019.112883