Research map / Graph learning / Graph self-supervised learning
A Persistent Weisfeiler-Lehman Procedure for Graph Classification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 cycle features (P-WL to P-WL-C) (ablation) | P-WL · MUTAG | accuracy: 86.10 ± 1.37 -> 90.51 ± 1.34 | Helped |
| Add cycle features (P-WL to P-WL-C) (ablation) | P-WL · PTC-MR | accuracy: 63.07 ± 1.68 -> 64.02 ± 0.82 | Helped |
| Add cycle features (P-WL to P-WL-C) (ablation) | P-WL · D & D | accuracy: 79.34 ± 0.46 -> 78.66 ± 0.32 | Hurt |
| Persistent-only features (P-WL) instead of WL subtree features (ablation) | WL · PTC-MM | accuracy: 67.28 ± 0.97 -> 68.40 ± 1.17 | Helped |
| Persistent-only features (P-WL) instead of WL subtree features (ablation) | WL · NCI1 | accuracy: 85.58 ± 0.15 -> 85.34 ± 0.14 | No clear effect |
| Uniform-weighted cycle variant (P-WL-UC) vs WL (ablation) | WL · PTC-FM | accuracy: 64.80 ± 0.85 -> 65.44 ± 1.18 | Helped |
| Persistent features instead of WL features, varying h (ablation) | WL · PTC-MR | accuracy vs h | Helped |
| Remove node labels; use degree propagation P-WL-D (h<=5) (ablation) | P-WL-D · PROTEINS | accuracy: 76.11 ± 0.64 -> 73.31 ± 0.53 | Hurt |
| P-WL-D with node degrees only (h=0) vs RETGK | RETGK · IMDB-BINARY | accuracy: 72.3 ± 0.6 -> 72.88 ± 0.55 | Helped |
| P-WL-UC with uniform labels vs RETGK | RETGK · IMDB-BINARY | accuracy: 72.3 ± 0.6 -> 73.02 ± 0.95 | Helped |
| Persistent features vs RETGK | RETGK · PTC data sets | accuracy | Helped |
| Change sigma in P-WL-D (ablation) | P-WL-D · graph benchmarks | accuracy | No clear effect |
Closest papers by meaning
- On topological descriptors for graph products (NeurIPS 2025 · similarity 0.88)
- Graph Persistence goes Spectral (NeurIPS 2025 · similarity 0.88)
- Generalizing Weisfeiler-Lehman Kernels to Subgraphs (ICLR 2025 · similarity 0.88)
- Structural Entropy Based Graph Structure Learning for Node Classification (AAAI 2024 · similarity 0.86)
- Graphcode: Learning from multiparameter persistent homology using graph neural networks (NeurIPS 2024 · similarity 0.86)
- Weisfeiler and Lehman Go Paths: Learning Topological Features via Path Complexes (AAAI 2024 · 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.