Bridging Academia and Industry: A Comprehensive Benchmark for Attributed Graph Clustering
PyAGC is a production-ready benchmark and library for Attributed Graph Clustering (AGC). It unifies existing AGC methods into an Encode-Cluster-Optimize framework, providing memory-efficient mini-batch implementations. The benchmark tests algorithms across 12 diverse industrial datasets ranging from 2.7K to 111M nodes, establishing a new standard for realistic deployment.