Alleviating Limited Effectiveness and Multi-Class Collapse in Deep Graph-level Clustering

Xiaoyang Lin, Renchi Yang, Yuqi Chen, Xin Huang

Given a collection of $N$ graphs, graph-level clustering (GLC) seeks to partition the $N$ graphs into semantically meaningful groups, which serves as a fundamental task in analyzing graph-structured data. While Graph Isomorphism Networks (GINs) have become the de facto encoders for GLC, they still fail to distinguish graphs due to their neglect of global topology. On top of that, existing deep clustering frameworks suffer from severe "clustering collapse" issues, particularly in imbalanced multi-class scenarios. To address these challenges, we propose S$^3$GLC, a novel framework that integrates global topological characteristics with local neighborhood features for balanced GLC. Theoretically, we leverage the Hoffman-Wielandt inequality to construct Spectral Signature as a rigorous lower bound for graph edit distance (GED), providing a computationally efficient proxy for global topological features. We additionally propose a joint manifold-preserving sparse regression module to fuse Spectral Signature with Spatial Representation, eliminating noise while preserving the complementary information. Furthermore, to mitigate clustering collapse, we introduce a balanced clustering…