Network embedding based on structural information entropy: Network embedding

Abstract

Blockchain is a complex network structure, which has the properties of members and the network relationship of their transactions with each other. If we can quantify its network structure, it will be of great help to our subsequent work. Network embedding is a method of analyzing a network to learn the low-dimensional potential representation of vertices in a continuous vector space, and the representation results will be easily applied to various statistical models and machine learning algorithms. Previous studies have focused on preserving the structural information of vertices at specific scales. Inspired by entropy, we propose a method of network embedding based on the structure information entropy of net-work, which can embed multiple layers of network and interact between layers. The federated node is embedded with the community. Experiments show that our method can well retain the structural in-formation of the network, and also can well show the connections between nodes and communities. In order to explicitly maintain the hierarchy of the network, we embed not only the vertices of the network, but also the community of all layers of the network.