Research map / Graph learning / GNN architectures
Connectivity-Guided Sparsification of 2-FWL GNNs: Preserving Full Expressivity with Improved Efficiency
Rongqin Chen, Fan Mo, Pak Lon Ip, Shenghui Zhang, Dan Wu, Ye Li, Leong Hou U
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Co-Sparsify connectivity-guided sparsification to PPGN+RRWP | PPGN+RRWP · ZINC-subset | test MAE: 0.055 -> 0.050 | Improved |
| add Co-Sparsify to PPGN+RRWP | PPGN+RRWP · ZINC-Full | test MAE: 0.020 -> 0.018 | Improved |
| add Co-Sparsify to PPGN+RRWP | PPGN+RRWP · QM9 U0 | MAE: 0.0111 -> 0.0149 | Hurt |
| add Co-Sparsify to PPGN+RRWP | PPGN+RRWP · QM9 G | MAE: 0.01278 -> 0.01767 | Hurt |
| add Co-Sparsify to PPGN+RRWP | PPGN+RRWP · QM9 H | MAE: 0.01266 -> 0.01637 | Hurt |
| add Co-Sparsify to PPGN+RRWP | PPGN+RRWP · QM9 Cv | MAE: 0.08589 -> 0.07619 | Improved |
| add Co-Sparsify to PPGN | PPGN · Substructure counting (8 targets) | normalized MAE: 3-Cycles 0.0003 -> 0.0004; Tri.-Rec. 0.0144 -> 0.0049 | Mixed |
| add Co-Sparsify to PPGN (TUD classification vs UnionGNNs) | PPGN · TUD FRANKENSTEIN | accuracy (%): 68.02 -> 77.65 | Improved |
| add Co-Sparsify to PPGN | PPGN · ZINC-subset | runtime per epoch (s): 9.3s -> 7.9s | Improved |
| add Co-Sparsify to PPGN | PPGN · QM9 | runtime per epoch (s): 97.1s -> 60.7s | Improved |
| add Co-Sparsify to PPGN | PPGN · QM9 | memory (GB): 6.4GB -> 4.2GB | Improved |
| use standard CoSp-PPGN on long-range task | CoSp-PPGN · Peptides-struct | MAE | Hurt |
Closest papers by meaning
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (ICML 2024 · similarity 0.89)
- Unifews: You Need Fewer Operations for Efficient Graph Neural Networks (ICML 2025 · similarity 0.87)
- Chaining 2-FWL GNNs for Combinatorial Graph Alignment (NeurIPS 2026 · similarity 0.87)
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.87)
- EAGLES: Towards Effective, Efficient, and Economical Federated Graph Learning via Unified Sparsification (ICML 2025 · similarity 0.87)
- What Expressivity Theory Misses: Message Passing Complexity for GNNs (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
