Research map / Architectures and efficiency / CNN design
B-Spline CNNs on Lie groups
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 an SE(2) G-CNN instead of a 2D CNN | Tumor classification · PCam | classification performance | Helped |
| Increase rotation sampling resolution (ablation) | SE(2) G-CNN · PCam | classification performance | Mixed |
| Use atrous kernels instead of dense kernels (ablation) | SE(2) G-CNN · PCam | classification performance | Helped |
| Use localized kernels with low N_k instead of dense kernels (ablation) | SE(2) G-CNN · PCam | classification performance | Helped |
| Remove 90° rotation augmentation (ablation) | 2D CNN · PCam | classification performance | Hurt |
| Remove 90° rotation augmentation (ablation) | SE(2) G-CNN · PCam | classification performance | No clear effect |
| Use a scale-translation G-CNN instead of a fixed-scale 2D CNN | Facial landmark localization · CelebA | landmark localization performance | Helped |
| Use localized kernels that interact between neighboring scales (ablation) | Scale-translation G-CNN · CelebA | landmark localization performance | Helped |
| Replace neighboring-scale interactions with all-scale interactions (ablation) | Scale-translation G-CNN · CelebA | landmark localization performance | Hurt |
| Remove scale interactions between neighboring scales (ablation) | Scale-translation G-CNN · CelebA | landmark localization performance | Hurt |
| Add random scale augmentation (ablation) | 2D CNN · CelebA | landmark localization performance | Helped |
| Increase scale augmentation beyond a factor of 1.4 (ablation) | 2D CNN · CelebA | landmark localization performance | No clear effect |
Closest papers by meaning
- Exploiting Redundancy: Separable Group Convolutional Networks on Lie Groups (ICML 2022 · similarity 0.90)
- Lie Group Decompositions for Equivariant Neural Networks (ICLR 2024 · similarity 0.89)
- Variational Partial Group Convolutions for Input-Aware Partial Equivariance of Rotations and Color-Shifts (ICML 2024 · similarity 0.89)
- Group Downsampling with Equivariant Anti-aliasing (ICLR 2025 · similarity 0.87)
- Continuous Rotation Group Equivariant Network Inspired by Neural Population Coding (AAAI 2024 · similarity 0.86)
- On the VC dimension of deep group convolutional neural networks (NeurIPS 2025 · similarity 0.85)
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.