Research map / Graph learning / GNN architectures
Provably expressive temporal graph networks
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 relative positional features to PINT (ablation) | PINT · UCI transductive | AP: 92.68 ± 0.5 -> 96.01 ± 0.1 | Helped |
| Add relative positional features to PINT (ablation) | PINT · Enron transductive | AP: 83.06 ± 2.1 -> 88.71 ± 1.3 | Mixed |
| Add relative positional features to PINT (ablation) | PINT · LastFM transductive | AP: 81.35 ± 1.6 -> 88.06 ± 0.7 | Helped |
| Add relative positional features to PINT (ablation) | PINT · Enron inductive | AP: 75.99 ± 2.3 -> 81.05 ± 2.4 | Mixed |
| Add relative positional features to TGN-Att | TGN-Att · UCI transductive | AP: 80.40 ± 1.4 -> 95.64 ± 0.1 | Helped |
| Add relative positional features to TGN-Att | TGN-Att · Enron inductive | AP: 78.96 ± 0.5 -> 76.27 ± 3.4 | Hurt |
| Increase relative positional feature dimension to d=10 (ablation) | PINT · Enron transductive | AP: 92.69 ± 0.09 | Helped |
| Increase relative positional feature dimension to d=10 (ablation) | PINT · Enron inductive | AP: 88.34 ± 0.29 | Helped |
| Increase relative positional feature dimension to d=20 (ablation) | PINT · UCI transductive | AP: 96.36 ± 0.07 | Helped |
| Increase relative positional feature dimension to d=20 (ablation) | PINT · UCI inductive | AP: 94.77 ± 0.12 | Helped |
| Use injective aggregation and update functions | MP-TGN · Temporal node distinguishability | expressive power | Helped |
| Replace recurrent memory with a deeper memoryless MP-TGN, adding Δ layers | MP-TGN · Temporal node distinguishability | expressive power | No clear effect |
Closest papers by meaning
- Expressive Power of Temporal Message Passing (AAAI 2025 · similarity 0.90)
- State Space Models on Temporal Graphs: A First-Principles Study (NeurIPS 2024 · similarity 0.89)
- PRES: Toward Scalable Memory-Based Dynamic Graph Neural Networks (ICLR 2024 · similarity 0.88)
- The Logical Expressiveness of Temporal GNNs via Two-Dimensional Product Logics (NeurIPS 2025 · similarity 0.87)
- Improving Temporal Link Prediction via Temporal Walk Matrix Projection (NeurIPS 2024 · similarity 0.87)
- Future Link Prediction Without Memory or Aggregation (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.