Research map / Point clouds / Point cloud classification
Invariant and Equivariant Reynolds Networks
Akiyoshi Sannai, Makoto Kawano, Wataru Kumagai
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 |
|---|---|---|---|
| use Corner MSE loss instead of standard MSE loss (ablation) | 4-red ReyNet · Synthetic equivariant (n=5) | MSE: 4.912e-5 -> 9.375e-5 | Hurt |
| use Corner MSE loss instead of standard MSE loss (ablation) | 4-red ReyNet · Synthetic equivariant (n=3) | MSE: 1.438e-4 -> 1.249e-4 | Improved |
| use Corner MSE loss instead of standard MSE loss (ablation) | 4-red ReyNet · Synthetic equivariant (n=20) | MSE: 1.608e-4 -> 8.537e-5 | Improved |
| add 4-reduction to ReyNet | ReyNet · Synthetic equivariant Symmetry (n=20) | MSE: 3.151e-3 -> 6.554e-5 | Improved |
| replace IEGN with 4-red ReyNet | IEGN · Synthetic equivariant Diagonal (n=20) | MSE: 4.743e-4 -> 3.566e-5 | Improved |
| train on n=3 and test on larger input sizes | 4-red ReyNet · Synthetic Symmetry extrapolation | MSE | Improved |
| test invariant IEGN on larger input sizes | IEGN · invariant tasks extrapolation | generalization | Hurt |
| use 4-red ReyNet instead of IEGN | IEGN · MUTAG | accuracy: 78.33 ± 10 -> 89.44 ± 7.1 | Improved |
| use 4-red ReyNet-(i) instead of GIN | GIN · PTC | accuracy: 56.64 ± 7.0 -> 61.18 ± 5.2 | Improved |
| use 4-red ReyNet-(i) instead of IEGN | IEGN · COLLAB | accuracy: 78.12 ± 2.9 -> 73.62 ± 1.6 | Hurt |
| use larger hidden sizes (128, 256, 512) instead of (16, 16, 16) (ablation) | 4-red ReyNet · NCI1 | accuracy: 77.25 ± 2.0 -> 77.79 ± 2.1 | No clear effect |
| use ReyNet instead of GIN | GIN · MOL-HIV | accuracy: 0.7366 ± 0.0268 -> 0.7342 ± 0.0161 | No clear effect |
Closest papers by meaning
- A Tale of Two Symmetries: Exploring the Loss Landscape of Equivariant Models (NeurIPS 2025 · similarity 0.87)
- On the hardness of learning under symmetries (ICLR 2024 · similarity 0.87)
- MatrixNet: Learning over symmetry groups using learned group representations (NeurIPS 2024 · similarity 0.87)
- Lie Group Decompositions for Equivariant Neural Networks (ICLR 2024 · similarity 0.87)
- Flow Equivariant Recurrent Neural Networks (NeurIPS 2025 · similarity 0.86)
- On Universality Classes of Equivariant 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.
