Abstract
Graphs representation learning has been a very active research area in recentyears. The goal of graph representation learning is to generate graphrepresentation vectors that capture the structure and features of large graphsaccurately. This is especially important because the quality of the graphrepresentation vectors will affect the performance of these vectors indownstream tasks such as node classification, link prediction and anomalydetection. Many techniques are proposed for generating effective graphrepresentation vectors. Two of the most prevalent categories of graphrepresentation learning are graph embedding methods without using graph neuralnets (GNN), which we denote as non-GNN based graph embedding methods, and graphneural nets (GNN) based methods. Non-GNN graph embedding methods are based ontechniques such as random walks, temporal point processes and neural networklearning methods. GNN-based methods, on the other hand, are the application ofdeep learning on graph data. In this survey, we provide an overview of thesetwo categories and cover the current state-of-the-art methods for both staticand dynamic graphs. Finally, we explore some open and ongoing researchdirections for future work.