Research map / Learning theory and algorithms / Matrix and tensor methods
Fast and Stable Riemannian Metrics on SPD Manifolds via Cholesky Product Geometry
Ziheng Chen, Yue Song, Xiaojun Wu, Nicu Sebe
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 CDEM instead of LCM in SPD MLR | SPDNet · FPHA | accuracy: 86.37 -> 89.40 | Improved |
| use CDBWM instead of LCM in SPD MLR | SPDNet · HDM05 3-block | accuracy: 62.33 -> 67.40 | Improved |
| use CDBWM instead of AIM in SPD MLR | SPDNet · Radar | accuracy: 94.53 -> 93.93 | Hurt |
| use CDEM instead of LCM in SPD MLR | GyroSPD · HDM05 | accuracy: 68.37 -> 71.93 | Improved |
| use CDBWM instead of AIM in SPD MLR | GyroSPD · Radar | accuracy: 96.80 -> 96.21 | Hurt |
| use CDEM instead of LCM in residual block | RResNet · HDM05 | accuracy: 66.27 -> 68.05 | Improved |
| tune Cholesky power deformation theta (-0.5) instead of theta=1 (ablation) | CDEM on SPDNet · HDM05 3-block | accuracy: 64.58 -> 65.75 | Improved |
| tune Cholesky power deformation theta (-1.5) instead of theta=1 (ablation) | CDBWM on SPDNet · HDM05 3-block | accuracy: 63.33 -> 67.40 | Improved |
| tune Cholesky power deformation theta (-0.75) instead of theta=1 (ablation) | CDEM on GyroSPD · HDM05 | accuracy: 67.80 -> 71.93 | Improved |
| set extreme theta=-2 in CDEM (ablation) | CDEM on SPDNet · HDM05 3-block | accuracy: 64.58 -> 46.63 | Hurt |
| tune theta (-1.5) instead of theta=1 (ablation) | CDEM on SPDNet · Radar | accuracy: 93.87 -> 95.79 | Improved |
| use diagonal power deformation (DPM, theta=0.15) instead of diagonal log (DLM) (ablation) | geodesic computation · 3x3 Cholesky, eps=1e-3 | failure probability (%): 51.32 -> 0 | Improved |
Closest papers by meaning
- RMLR: Extending Multinomial Logistic Regression into General Geometries (NeurIPS 2024 · similarity 0.91)
- Riemannian Multinomial Logistics Regression for SPD Neural Networks (CVPR 2024 · similarity 0.91)
- Learning to Normalize on the SPD Manifold under Bures-Wasserstein Geometry (CVPR 2025 · similarity 0.88)
- Matrix Manifold Neural Networks++ (ICLR 2024 · similarity 0.86)
- Neural networks on Symmetric Spaces of Noncompact Type (ICLR 2025 · similarity 0.86)
- Understanding Matrix Function Normalizations in Covariance Pooling through the Lens of Riemannian Geometry (ICLR 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.
