Graph Clustering

Graph clustering studies how to divide network nodes into groups, helping reveal communities, organize large graphs, and support tasks such as anomaly detection and graph learning. Classical approaches use spectral methods, random-walk embeddings, modularity optimization, and matrix factorization, while newer methods combine graph neural networks, node attributes, diffusion, and optimal transport. The field also develops statistical guarantees for stochastic block models and approximation algorithms for constrained or hierarchical partitions. A central challenge is making methods scale while remaining reliable on noisy, directed, temporal, or heterophilous graphs, where standard assumptions and evaluation metrics may not reflect meaningful groups.