Reproduce “Graph Distillation with Eigenbasis Matching” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yang Liu, Deyu Bo, Chuan Shi

Researchers from Beijing University of Posts and Telecommunication developed the Graph Distillation with Eigenbasis Matching (GDEM) framework, which synthesizes smaller graphs by aligning the eigenbasis and replicating the spectrum of original graphs. This method achieves better generalization across diverse Graph Neural Network architectures and substantially reduces distillation time compared to prior GNN-dependent approaches.

Key insights

  • GNN-dependent graph distillation methods inherently introduce 'spectrum bias' (favoring certain frequency components) and necessitate costly architectural 'traversal' due to their reliance on specific GNN filtering characteristics.
  • Directly preserving intrinsic spectral properties, specifically the eigenbasis, can decouple graph distillation from specific GNN architectures, leading to better cross-architecture generalization.
  • Achieving optimal graph distillation involves balancing the preservation of general structural information (via eigenbasis matching) with task-specific category-level discrimination.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.