Signed Network Representation Learning Algorithm Based on Structural Balance Theory and High-Order Mutual Information

In this paper, a symbolic network representation learning framework signed network representation learning algorithm based on structural balance theory and high-order mutual information (SNSH) is proposed. By reversing the positive and negative relationships in symbolic networks to generate negative graphs, the hidden high-order mutual information in symbolic networks is mined. This method aims to simulate the local implicit features of symbolic networks through the strengthened social balance theory, and obtain a more comprehensive node embedding that conforms to the characteristics of symbolic networks through the high-order mutual information among the local embedding of nodes, the global structure of the network and the characteristic attributes of nodes.