Introduction to Graph Machine Learning

This post introduces the fundamentals of graph machine learning, covering graph definitions (nodes, edges, directed/undirected, homogeneous/heterogeneous), common tasks (graph-level, node-level, edge-level prediction), and representation methods (adjacency matrix, edge lists). It details pre-neural approaches like feature engineering (centrality, degree, clustering coefficient, graphlets) and walk-based methods. It then introduces Graph Neural Networks (GNNs), explaining aggregation and message passing, and the over-smoothing problem. Finally, it briefly touches upon Graph Transformers and provides further resources.