Research map / 3D vision / Depth and stereo
Eigendecomposition-Free Training of Deep Networks for Linear Least-Square Problems
Zheng Dang, Kwang Moo Yi, Yinlin Hu, Fei Wang, Pascal Fua, Mathieu Salzmann
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 |
|---|---|---|---|
| replace analytical SVD/Eigh gradients with the eigendecomposition-free loss | plane fitting with a single outlier · Plane fitting (single outlier) | loss decrease during optimization | Improved |
| replace analytical SVD/Eigh gradients with the eigendecomposition-free loss | plane fitting with twenty outliers · Plane fitting (20 outliers, Adam) | convergence: convergence suddenly happens after 352 iterations for SVD/Eigh | Improved |
| add the eigendecomposition-free loss instead of Essential_Only eigendecomposition loss (ablation) | wide-baseline stereo keypoint matching · SUN3D wide-baseline stereo | mAP | Improved |
| drop the classification pre-training stage (ablation) | eigendecomposition-free loss · SUN3D wide-baseline stereo | mAP | No clear effect |
| use eigendecomposition-free loss instead of REPPnP | PnP outlier removal · PnP synthetic, outliers 10 to 150 | rotation error: more than 40 outliers, that is, 20% | Improved |
| apply synthetic-trained model to real data without fine-tuning | PnP outlier removal · PnP real data (1000 images) | reprojection accuracy | Improved |
| use eigendecomposition-free loss under combined noise and outliers | PnP simultaneous denoising and outlier removal · PnP synthetic noise+outliers | rotation and translation error | Improved |
| use eigendecomposition-free loss for denoising | ellipse fitting · Ellipse fitting denoising | center point error | Improved |
Closest papers by meaning
- Eigendecomposition-free Training of Deep Networks with Zero Eigenvalue-based Losses (ECCV 2018 · similarity 0.97)
- Amortized Eigendecomposition for Neural Networks (NeurIPS 2024 · similarity 0.88)
- Schur's Positive-Definite Network: Deep Learning in the SPD cone with structure (ICLR 2025 · similarity 0.86)
- Preserving Deep Representations in One-Shot Pruning: A Hessian-Free Second-Order Optimization Framework (ICLR 2025 · similarity 0.85)
- STNet: Spectral Transformation Network for Solving Operator Eigenvalue Problem (NeurIPS 2025 · similarity 0.85)
- Eigenspectrum Analysis of Neural Networks without Aspect Ratio Bias (ICML 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.
