Research map / Graph learning / Graph algorithms and clustering
SeedGNN: Graph Neural Network for Supervised Seeded Graph Matching
Liren Yu, Jiaming Xu, Xiaojun Lin
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 |
|---|---|---|---|
| Pair-wise SeedGNN with seeds encoded as 0/1 indicators vs node-based semi-supervised / supervised GNNs | seeded graph matching · SHREC'16 | Accuracy (%): 43.1 vs 37.9 (best baseline NGM) | Improved |
| Hungarian-based cleaning (masking) of similarity matrix before passing seeds to next layer (ablation) | MLP+SeedGNN pipeline on Willow Object · Willow Object | Mean accuracy (%): 97.5 vs 85.6 | Improved |
| SeedGNN refinement of seedless DGMC output | DGMC on Willow Object · Willow Object | Mean accuracy (%): 98.9 vs 93.7 | Improved |
| SeedGNN refinement of MLP similarities (no cleaning) | MLP on Willow Object · Willow Object | Mean accuracy (%): 85.6 vs 71.1 | Improved |
| Pair-wise architecture normalization removed (SeedGNN vs NGM-like normalization by degrees) | correlated Erdos-Renyi test graphs with larger n and dense p=0.2 · correlated Erdos-Renyi (n=500, p=0.2) | matching accuracy | Improved |
| Percolation module removed (SeedGNN-w) (ablation) | SeedGNN · correlated Erdos-Renyi (n=500, p=0.04) | matching accuracy | Improved |
| Convolution at larger hops removed (SeedGNN-x, 1-hop only) (ablation) | SeedGNN · correlated Erdos-Renyi (n=500, p=0.04) | matching accuracy | Improved |
| Hungarian cleaning removed from percolation (soft similarity only, SeedGNN-p) (ablation) | SeedGNN · correlated Erdos-Renyi (n=500, p=0.04) | matching accuracy | Improved |
| Using only hard Hungarian matches as new seeds without confidence (SeedGNN-h) (ablation) | SeedGNN · correlated Erdos-Renyi (n=500, p=0.04) | matching accuracy | Improved |
| Trained with NGM normalization vs not on SHREC'16 (NGM trained on SHREC'16 test-like degrees) | NGM on SHREC'16 · SHREC'16 | matching accuracy: NGM 37.9 vs SeedGNN 43.1 | No clear effect |
| Semi-supervised GNNs (DeepLink, CrossMNA, MGCN) vs supervised SeedGNN on SHREC'16 with few seeds | semi-supervised GNNs · SHREC'16 | Accuracy (%): 3.3-4.2 vs 43.1 | Hurt |
Closest papers by meaning
- Learning To Match Features With Seeded Graph Matching Network (ICCV 2021 · similarity 0.92)
- Graph Matching with Partially-Correct Seeds (JMLR 2021 · similarity 0.90)
- Combinatorial Learning of Robust Deep Graph Matching: An Embedding Based Approach (TPAMI 2020 · similarity 0.89)
- Chaining 2-FWL GNNs for Combinatorial Graph Alignment (NeurIPS 2026 · similarity 0.87)
- GraphTOP: Graph Topology-Oriented Prompting for Graph Neural Networks (NeurIPS 2025 · similarity 0.86)
- Charting the Design Space of Neural Graph Representations for Subgraph Matching (ICLR 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.
