Research map / Graph learning / GNN architectures
Geometrically Principled Connections in Graph Neural Networks
Check 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 affine skip connections to MoNet | MoNet · CoMA shape reconstruction, M=9 | mean Euclidean error (mm): 0.528 ± 0.604 -> 0.406 ± 0.455 | Helped |
| Add affine skip connections to FeaStNet | FeaStNet · CoMA shape reconstruction, M=9 | mean Euclidean error (mm): 0.524 ± 0.646 -> 0.470 ± 0.585 | Helped |
| Add affine skip connections to MoNet | MoNet · FAUST shape correspondence, geodesic error 0 | correct correspondences (%): 86.61% -> 94.69% | Helped |
| Add affine skip connections to MoNet | MoNet · Superpixel MNIST, vertex positions | accuracy (%): 96.57 -> 97.14 | Helped |
| Add affine skip connections to MoNet | MoNet · Superpixel MNIST, vertex degrees | accuracy (%): 53.10 -> 72.00 | Helped |
| Add affine skip connections to FeaStNet | FeaStNet · Superpixel MNIST | accuracy (%): 11.35 -> 59.50 | Helped |
| Add affine skip connections to GCN | GCN · Superpixel MNIST | accuracy (%): 31.21 -> 58.96 | Helped |
| Increase MoNet kernel size by one instead of adding affine skip connections | MoNet · CoMA shape reconstruction | mean Euclidean error (mm): 0.528 ± 0.604 -> 0.528 ± 0.587 | No clear effect |
| Remove the separate center-vertex weight (ablation) | ChebNet · CoMA shape reconstruction, M=9 | mean Euclidean error (mm): 0.438 ± 0.562 -> 4.329 ± 3.591 | Hurt |
| Remove the center vertex from the spiral filter (ablation) | SpiralNet++ · CoMA shape reconstruction, M=9 | mean Euclidean error (mm): 0.426 ± 0.538 -> 0.430 ± 0.542 | No clear effect |
| Remove self-loops (ablation) | FeaStNet · CoMA shape reconstruction, M=9 | mean Euclidean error (mm): 0.524 ± 0.646 -> 0.519 ± 0.634 | Helped |
| Remove self-loops while retaining affine skip connections (ablation) | Aff-FeaStNet · CoMA shape reconstruction, M=9 | mean Euclidean error (mm): 0.470 ± 0.585 -> 0.463 ± 0.577 | Helped |
Closest papers by meaning
- Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs (CVPR 2017 · similarity 0.89)
- Geometry Awakening: Cross-Geometry Learning Exhibits Superiority over Individual Structures (NeurIPS 2024 · similarity 0.87)
- What do Graph Neural Networks learn? Insights from Tropical Geometry (NeurIPS 2024 · similarity 0.87)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.86)
- On the Completeness of Invariant Geometric Deep Learning Models (ICLR 2025 · similarity 0.86)
- Geometry-Aware Edge Pooling for Graph Neural Networks (NeurIPS 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.