GraphSAGE

GraphSAGE is an inductive algorithm for computing node embeddings. GraphSAGE makes use of node feature information to generate node embeddings on unseen nodes or graphs. Instead of training individual embeddings for each node, the algorithm learns a function that generates embeddings by sampling and aggregating features from a node’s local neighborhood.

The algorithm is defined for UNDIRECTED graphs.

Node embeddings

The Neo4j Graph Data Science library contains two implementations of GraphSAGE for generating node embeddings.

GraphSAGE - The first implementation, available in the plugin as well as Aura Graph Analytics.

GraphSAGE unsupervised - The more recent implementation which offers greatly improved training speed. Only available on Aura Graph Analytics.

Node classification

GraphSAGE supervised - GraphSAGE can also be used as a supervised algorithm for node classification. Only available on Aura Graph Analytics.

Reference

For more in-depth information on this algorithm see:

Inductive Representation Learning on Large Graphs William L. Hamilton, Rex Ying, and Jure Leskovec. (2017)
SWAG: Item Recommendations using Convolutions on Weighted Graphs Amit Pande, Kai Ni and Venkataramani Kini. (2019)