A Tutorial of Graph Representation

Abstract

With the development of network technology, graphs exist widely in real life, such as social networks, communication networks, biological networks, etc. Analyzing the structural of these graphs will have a far-reaching impact on various network applications. For example, node classification, link prediction, clustering, visualization. Traditional graph representation methods suffer a lot of space and time cost. In recent years, a category of technology which convert nodes into a low-dimensional vector has emerged. In this paper, First, we briefly introduce the development and challenges of graph representation algorithms. Then we introduce the existing methods of graph representation in the literature. It is mainly divided into three parts: node embedding, embedding based on heterogeneous graph, subgraph embedding. Node embedding algorithms are mainly divided into four categories: matrix factorization-based algorithms, random walk based algorithms, deep learning based algorithms and the algorithm based on the role of node structure. In addition, we also introduce LINE algorithms, embedding based on heterogeneous graph and sub-graph (whole graph) embedding algorithms. Finally, we introduce the related applications of graph embedding and the summary of this paper.