Node Embedding Algorithms

Node embeddings are vector representations of properties of vertices in a graph. These vectors can be used for machine learning.

The TigerGraph Graph Data Science Library provides the following node embedding algorithms:

  • Fast Random Projection

  • NodePiece is a highly scalable node embedding technique. It is implemented in the pyTigerGraph library. Using it requires a few additional steps, but it is our most scalable offering.

  • Weisfeiler-Lehman isomorphism has been moved to the Classification category.

  • Node2Vec has been retired (as of v3.9.2) because it did not meet our standards for efficiency and scalability of memory usage. Fast Random Projection is a better alternative.