Link Prediction on Multiple Graphs with Graph Embedding and Optimal Transport

Link prediction is an extensively studied topic and various methods have been proposed to tackle the task in both heuristic and more sophisticated statistical learning approaches. However, most of them only focus on one single graph. In many scenarios, combining information on multiple graphs with similar topological structures can greatly improve the performance and robustness of link prediction. In this study, we propose a new framework for learning link prediction on two unaligned graphs simultaneously. We use the LINE method, although technically any embedding method is applicable, to embed nodes of each graph into low-dimensional vectors. Optimal Transport is then employed to supervise the node alignment via embedding vectors between the two graphs. The learned embedding vectors are employed for link prediction via a similarity score. Experiments have shown that node alignment using Optimal Transport is beneficial and greatly contributes to the favorable performance of the proposed method over the baseline in many settings.