Attributed Graph Embedding with Random Walk Regularization and Centrality-Based Attention

Graph-embedding learning is the foundation of complex information network analysis, aiming to represent nodes in a graph as low-dimensional dense real-valued vectors for application practical analysis tasks. In recent years, study representation has received increasing attention from researchers, and, among them, neural networks (GNNs) based on deep are playing an increasingly important role this field. However, fact that higher-order neighborhood cannot be used effectively problem most existing networks. Moreover, it tends ignore influence latent and structural properties embedding. hopes solving these issues, we introduce centrality encoding learn node properties, add mechanism consideration better distinguish significance neighboring nodes, random walk regularization make sample neighbors consistently satisfy predetermined criteria. This allows us potential node. We tested performance our model node-clustering link prediction tasks using three widely recognized benchmark datasets. The outcomes experiments demonstrate significantly surpasses baseline method both tasks, indicating embedding generates highly expressive.