Node Embedding,Link Prediction And Network Alignment In Networks

Lots of data cannot be represented in Euclidean space in machine learning,such as social networks,biological networks,etc..,which are often composed of points and edges,forming a topological graph.Based on the topological graph,many graph model algorithms emerge as the times require.They play an important role in today's machine learning stage and are widely used.Among them,link prediction and network alignment are two typical applications of graph model.The former predicts to find the potential or future connected edge in the unconnected edge according to the existing information;the latter obtains the nodes with the same components in the given two similar networks.Researches on these two tasks are well done,but few people think that they are related.Based on the principle of Skip-gram,this paper obtains the embedding vector of each node in two graphs,and explains the relationship between link prediction and network alignment.In this paper,based on the graph embedding,the corresponding link prediction and network alignment algorithms are proposed respectively.Through theoretical explanation and experimental results,it can be found that link prediction and network alignment can indeed promote each other,where significant results are achieved in the experiments.