Feature Extraction for Graphs

Heads up: I've structured the article similarly as in the Graph Representation Learning book by William L. Hamilton [1]. One of the simplest ways to capture information from graphs is to create individual features for each node. These features can capture information both from a close neighbourhood, and a more distant, K-hop neighbourhood using iterative methods. Node degree is a simple metric and can be defined as a number of edges incident to a node. This metric is often used as initialization of algorithms to generate more complex graph-level features such as Weisfeiler-Lehman Kernel.