Research map / Graph learning / GNN architectures
Boosting the Cycle Counting Power of Graph Neural Networks with I$^2$-GNNs
Yinan Huang, Xingang Peng, Jianzhu Ma, Muhan Zhang
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 an identifier for each neighbor of the root node | Subgraph MPNNs, forming I^2-GNN · SR25 | accuracy: 0% -> 100% | Improved |
| Add an identifier for each neighbor of the root node | Subgraph MPNNs, forming I^2-GNN · Synthetic | normalized MAE, 5-Cycle: 0.0402 -> 0.0028 | Improved |
| Add an identifier for each neighbor of the root node | Subgraph MPNNs, forming I^2-GNN · Synthetic | normalized MAE, 6-Cycle: 0.0439 -> 0.0082 | Improved |
| Add an identifier for each neighbor of the root node | Subgraph MPNNs, forming I^2-GNN · ChEMBL | normalized MAE, 6-Cycle: 0.0053 -> 0.0003 | Improved |
| Add an identifier for each neighbor of the root node | Subgraph MPNNs, forming I^2-GNN · Synthetic | normalized MAE, 4-Cycle: 0.0009 -> 0.0016 | Mixed |
| Use I^2-GNN instead of NGNN | Molecular property prediction · QM9 | MAE across targets: 21.6% improvement | Improved |
| Use I^2-GNN instead of Nested GIN | Molecular graph regression · ZINC-12K | MAE: 25% performance gains | Improved |
| Use I^2-GNN instead of Nested GIN | Molecular graph regression · ZINC-250K | MAE: 18% performance gains | Improved |
| Use I^2-GNN instead of Nested GIN | Molecular graph classification · ogbg-molhiv | AUC: 0.3% | Improved |
| Use I^2-GNN instead of CIN | Molecular graph classification · ogbg-molhiv | AUC: $80.94\!\pm\!0.57$ -> $78.68\!\pm\!0.93$ | Hurt |
Closest papers by meaning
- Homomorphism Counts for Graph Neural Networks: All About That Basis (ICML 2024 · similarity 0.90)
- On the Scalability of GNNs for Molecular Graphs (NeurIPS 2024 · similarity 0.88)
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.88)
- Beyond Message Passing: Neural Graph Pattern Machine (ICML 2025 · similarity 0.87)
- Deep Homomorphism Networks (NeurIPS 2024 · similarity 0.86)
- Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality (ICML 2025 · similarity 0.86)
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.
