GraphSAGE

A GNN method that makes representation learning scalable on large graphs through neighborhood sampling.

GraphSAGE makes learning on large graphs more practical by aggregating sampled neighbors instead of requiring the full neighborhood structure. This also makes it possible to generalize learned representations to unseen nodes rather than only a fixed graph. That property is highly valuable in industrial-scale graph problems. It stands as an important example of scalable GNN design for large social, commercial, and interaction networks.