This talk will walk through a node embedding algorithm called HashGNN. Node embedding algorithms provide an important bridge between graphs and traditional machine learning. Such algorithms construct for each node in a graph an associated vector which can capture local graph structure and properties. We will discuss why HashGNN is a useful embedding algorithm, its implementation in GDS which is an extension to heterogenous graphs, some intuition behind it, how it compares to Graph Neural Networks which inspired HashGNN, and more.
Speakers: Adam Schill Collberg Jacob Sznajdman