Graph Sampling

Graph sampling refers to the process of selecting representative subgraph samples from large graph datasets. Its primary goal is to reduce computational complexity and resource consumption, thereby improving the efficiency of training Graph Neural Networks (GNNs) and generating graph embeddings. Graph sampling ensures that the selected samples retain key structures and properties of the original graph, thus maintaining model performance while accelerating algorithm convergence and optimization. In large-scale graph data processing, graph sampling techniques are of significant value for enhancing the scalability and practicality of model training.