Research map / Architectures and efficiency / CNN design
Riemannian Local Mechanism for SPD Neural Networks
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 |
|---|---|---|---|
| add multi-scale local submanifold features to SPDNet | SPDNet · FPHA | accuracy: 85.57 -> 87.13 | Helped |
| add multi-scale local submanifold features to SPDNet | SPDNet · CG | accuracy: 89.03 -> 91.25 | Helped |
| add multi-scale local submanifold features to SPDNet | SPDNet · UCF-sub | accuracy: 59.93 -> 60.87 | Helped |
| add selected proper local submanifolds to a holistic-only model (ablation) | MSNet-H · FPHA | accuracy: 85.74 -> 87.13 | Helped |
| add selected proper local submanifolds to a holistic-only model (ablation) | MSNet-H · CG | accuracy: 89.03 -> 91.25 | Helped |
| add selected proper local submanifolds to a holistic-only model (ablation) | MSNet-H · UCF-sub | accuracy: 58.27 -> 60.87 | Helped |
| add holistic/global information alongside proper submanifolds (ablation) | MSNet-S · FPHA | accuracy: 86.61 -> 87.13 | Helped |
| add holistic/global information alongside proper submanifolds (ablation) | MSNet-S · CG | accuracy: 90.14 -> 91.25 | Helped |
| add holistic/global information alongside proper submanifolds (ablation) | MSNet-S · UCF-sub | accuracy: 59.40 -> 60.87 | Helped |
| select all proper submanifolds instead of the task-specific selection (ablation) | MSNet · FPHA | accuracy: 87.13 -> 80.52 | Hurt |
| extract every submanifold, including trivial ones, rather than the selected subset (ablation) | MSNet · UCF-sub | accuracy: 60.87 -> 58.33 | Hurt |
| use backbone configuration {100,80,50,25,9} rather than {100,60,36} (ablation) | MSNet · UCF-sub | accuracy: 57.33 vs 36.07; SPDNet 56.73 vs 28.73 | Mixed |
Closest papers by meaning
- A Riemannian Network for SPD Matrix Learning (AAAI 2017 · similarity 0.92)
- Building Transformation Layers for Riemannian Neural Networks (NeurIPS 2026 · similarity 0.90)
- Matrix Manifold Neural Networks++ (ICLR 2024 · similarity 0.88)
- Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves (NeurIPS 2026 · similarity 0.88)
- Learning to Normalize on the SPD Manifold under Bures-Wasserstein Geometry (CVPR 2025 · similarity 0.87)
- Understanding Matrix Function Normalizations in Covariance Pooling through the Lens of Riemannian Geometry (ICLR 2025 · similarity 0.87)
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.