Research map / Graph learning / GNN architectures
SplineCNN: Fast Geometric Deep Learning With Continuous B-Spline Kernels
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 |
|---|---|---|---|
| use SplineCNN instead of MoNet | MNIST superpixel classification · MNIST superpixels | classification accuracy: 91.11% -> 95.22%; 4.11 percentage points | Helped |
| use SplineCNN instead of LeNet5 or MoNet | MNIST grid classification · MNIST grid | classification accuracy: 99.33% vs 99.22%; 99.19% vs 99.22% | No clear effect |
| use SplineCNN instead of the strongest listed baseline, CayleyNet | Cora node classification · Cora | accuracy: 87.90 ± 0.66 -> 89.48 ± 0.31; approximately 1.58 percentage points | Helped |
| use SplineCNN for correspondence prediction | FAUST shape correspondence · FAUST test | zero geodesic error predictions: 99.20 % | Helped |
| use SplineCNN with cross-entropy loss rather than FMNet's specialized soft error loss | FAUST shape correspondence · FAUST test | accuracy at larger geodesic error bounds | Hurt |
| use Cartesian coordinates and a lower B-spline degree (ablation) | SplineCNN · MNIST superpixels | classification accuracy | Helped |
| use Cartesian coordinates and a lower B-spline degree (ablation) | SplineCNN · FAUST shape correspondence | geodesic correspondence accuracy | Helped |
| increase network depth beyond the task's six convolutional layers (ablation) | SplineCNN · FAUST shape correspondence | accuracy | No clear effect |
Closest papers by meaning
- SplineNets: Continuous Neural Decision Graphs (NeurIPS 2018 · similarity 0.91)
- FeaStNet: Feature-Steered Graph Convolutions for 3D Shape Analysis (CVPR 2018 · similarity 0.88)
- Geometrically Principled Connections in Graph Neural Networks (CVPR 2020 · similarity 0.88)
- Primal-Dual Mesh Convolutional Neural Networks (NeurIPS 2020 · similarity 0.88)
- Learning Convolutional Neural Networks for Graphs (ICML 2016 · similarity 0.88)
- Convolutional Kernel Networks for Graph-Structured Data (ICML 2020 · similarity 0.88)
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.