Research map / Architectures and efficiency / CNN design
Relaxing Equivariance Constraints with Non-stationary Continuous Filters
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 |
|---|---|---|---|
| Relax rotation equivariance with a cross-validated non-stationary filter | SE(2)-CNN with 8 rotation samples · CIFAR-10, no augment | test accuracy: 84.82 -> 86.39 | Helped |
| Relax rotation equivariance with a cross-validated non-stationary filter | SE(2)-CNN with 8 rotation samples · CIFAR-10, augment | test accuracy: 87.37 -> 88.38 | Helped |
| Relax rotation equivariance with a cross-validated non-stationary filter | SE(2)-CNN with 8 rotation samples · CIFAR-100, no augment | test accuracy: 51.04 -> 54.68 | Helped |
| Relax rotation equivariance with a cross-validated non-stationary filter | SE(2)-CNN with 8 rotation samples · CIFAR-100, augment | test accuracy: 58.22 -> 60.83 | Helped |
| Relax translation equivariance | T(2)-CNN · CIFAR-10 and CIFAR-100 | test accuracy: 81.80 -> 81.82; 85.87 -> 87.74; 46.20 -> 46.84; 53.97 -> 58.12 | Mixed |
| Relax rotation equivariance | G-CNN · MNIST6-180 | test accuracy: 50.0 -> 100.0 | Helped |
| Learn the relaxation frequency with gradient updates and symmetry regularization (ablation) | Soft SE(2)-CNN with 8 rotation samples · CIFAR-10 and CIFAR-100 | test accuracy: 86.39 -> 86.36; 88.38 -> 89.26; 54.68 -> 54.41; 60.83 -> 60.94 | Mixed |
| Use λ=0.001 or λ=0.01 for gradient-learned relaxation (ablation) | Soft SE(2)-CNN with 8 rotation samples · CIFAR-10 and CIFAR-100 | test accuracy: 86.39 -> 83.54; 88.38 -> 87.10; 54.68 -> 50.56; 60.83 -> 56.48 | Hurt |
| Use λ=0.5 for gradient-learned relaxation (ablation) | Soft SE(2)-CNN with 8 rotation samples · CIFAR-100, no augment | test accuracy: 54.68 -> 57.11 | Helped |
Closest papers by meaning
- Learning (Approximately) Equivariant Networks via Constrained Optimization (NeurIPS 2025 · similarity 0.92)
- A Tale of Two Symmetries: Exploring the Loss Landscape of Equivariant Models (NeurIPS 2025 · similarity 0.91)
- On Making $SE(2)$-Invariant Networks Optimal (NeurIPS 2026 · similarity 0.91)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.91)
- Improving Equivariant Model Training via Constraint Relaxation (NeurIPS 2024 · similarity 0.91)
- Probing Equivariance and Symmetry Breaking in Convolutional Networks (NeurIPS 2025 · similarity 0.90)
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.