Research map / Graph learning / GNN architectures
Pointer 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 pointer masking to PGN (ablation) | PGN · DSU n=100 ops=150 | F1: 0.499 -> 0.866 | Helped |
| add pointer masking to PGN (ablation) | PGN · LCT n=100 ops=150 | F1: 0.353 -> 0.616 | Helped |
| add pointer masking to PGN (ablation) | PGN · DSU n=20 ops=30 | F1: 0.910 -> 0.895 | Hurt |
| use pointer and mask supervision in the model (PGN vs SupGNN) (ablation) | GNN with extra supervision · LCT n=100 ops=150 | F1: 0.541 -> 0.616 | Helped |
| add extra supervision losses to GNN without using pointers (ablation) | GNN · LCT n=100 ops=150 | F1: 0.401 -> 0.541 | Helped |
| direct pointer supervision instead of DGM loss (ablation) | PGN · LCT n=100 ops=150 | F1: 0.524 -> 0.616 | Helped |
| add pointer supervision on top of mask-only supervision (ablation) | PGN-MO · LCT n=100 ops=150 | F1: 0.558 -> 0.616 | Helped |
| symmetrise pointers (ablation) | PGN · LCT n=100 ops=150 | F1: 0.561 -> 0.616 | Helped |
| replace unrestricted GNN with PGN | GNN · DSU n=100 ops=150 | F1: 0.733 -> 0.866 | Helped |
| replace unrestricted GNN with PGN | GNN · LCT n=100 ops=150 | F1: 0.401 -> 0.616 | Helped |
| replace learned PGN pointers with ground-truth pointers | PGN · LCT n=200 ops=300 | F1: 0.636 -> 0.619 | No clear effect |
Closest papers by meaning
- Graph Pointer Neural Networks (AAAI 2022 · similarity 0.90)
- Position-aware Graph Neural Networks (ICML 2019 · similarity 0.89)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.88)
- What functions can Graph Neural Networks compute on random graphs? The role of Positional Encoding (NeurIPS 2023 · similarity 0.88)
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.88)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · 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.