Local Structural Aware Heterogeneous Information Network Embedding Based on Relational Self-Attention Graph Neural Network

Meng Cao,Ming Xu,Hualei Yu,Chongjun Wang,Jinliang Yuan

doi:10.1109/access.2021.3090055

Meng Cao, Ming Xu + Show 3 more

Open Access

https://doi.org/10.1109/access.2021.3090055

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2021
Citations: 2	License type: CC BY 4.0

Affiliation: Nanjing University of Science and Technology

Abstract

Heterogeneous Information Networks (HINs) are ubiquitous in our daily life, as they can describe complex interactions among various types of objects. Heterogeneous information network embedding aims to project the network elements of HINs into low-dimensional node representation vectors, which can facilitate effective analyze of HINs. To learn appropriate HIN embeddings, most existing methods usually adopt meta-paths to capture the heterogeneous semantic information, which require domain knowledge and trial-and-error to find suitable meta-paths. Besides, most existing methods fail to preserve high-order local structural information, and pay little attention to the implicit semantics in HINs. In this paper, we propose a local structural aware heterogeneous information network embedding model named LSA-HNE. Specifically, we first design a relational self-attention graph neural network model to aggregate heterogeneous information and automatically extract semantic similarity without using meta-paths. In addition, we employ a biased random walk based sampling method to extract the local structural information and preserve the implicit semantics in HINs. The experiments conducted on four real-world datasets show that our proposed model is effective compared with the state-of-the-art methods.

Highlights

H ETEROGENEOUS Information Networks (HINs) [1] are ubiquitous in human society, which contain abundant information with multiple types of nodes and complex interactions, as well as rich attribute information
Such methods rely on manually designed meta-paths to extract semantic information, and leverage shallow neural network models for network embedding learning
With the purpose of extracting the implicit semantics and better exploiting the local structural information, we introduce the local structural aware embedding learning method, which comprises two parts, local structural information extraction based on biased random walk with restart, and local structural aware loss computing based on explicit and implicit edge sampling

Summary

Introduction

H ETEROGENEOUS Information Networks (HINs) [1] are ubiquitous in human society, which contain abundant information with multiple types of nodes and complex interactions, as well as rich attribute information. To facilitate effective representation learning on HINs, many HIN embedding models have been proposed, such as metapath2vec [5], HIN2VEC [6] and HERec [7]. Such methods rely on manually designed meta-paths to extract semantic information, and leverage shallow neural network models for network embedding learning.

Objectives

Methods

Findings

Conclusion