Research map / Learning theory and algorithms / Matrix and tensor methods
Generalized Shuffled Linear Regression
Feiran Li, Kent Fujiwara, Fumio Okura, Yasuyuki Matsushita
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 GSLR (outlier-rejecting one-to-one matching) instead of CPD | CPD · Synthetic point clouds with outliers | mean transformational error: 1.58e-2 -> 6.24e-3 | Improved |
| use GSLR instead of Trim-ICP | Trim-ICP · Synthetic point clouds with outliers | mean transformational error: 1.07 -> 6.24e-3 | Improved |
| use GSLR instead of ICP | ICP · Synthetic point clouds with partial overlap | mean transformational error: 1.70e-1 -> 6.52e-2 | Improved |
| use GSLR instead of CPD | CPD · Synthetic point clouds with partial overlap | median transformational error: 9.09e-2 -> 3.78e-2 | Improved |
| use GSLR instead of FGR | FGR · Synthetic point clouds with partial overlap | mean transformational error: 1.07 -> 6.52e-2 | Improved |
| use GSLR instead of ICP | ICP · ETH Hauptgebaude real-world point clouds | mean transformational error: 5.82e-1 -> 1.40e-1 | Improved |
| use GSLR instead of CPD | CPD · ETH Hauptgebaude real-world point clouds | median transformational error: 1.22e-1 -> 1.16e-1 | Improved |
| use GSLR instead of FGR | FGR · ETH Hauptgebaude real-world point clouds | mean transformational error: 1.23 -> 1.40e-1 | Improved |
| fix inlier count k to all points (SLR) vs estimating k via Huber-skip (GSLR) (ablation) | SLR · Image registration (SIFT points, homography) | registration quality: SLR distorted on all trials -> GSLR matches ground truth visually | Improved |
| use GSLR instead of ICP/Trim-ICP under coarse initialization | ICP and Trim-ICP · Bunny point clouds with 1/3 outliers | sensitivity to initialization: fail drastically when coarsened -> stable under large initialization errors | Improved |
| use GSLR instead of ZoomOut-100 for shape matching | ZoomOut-100 · FAUST/TOSCA shape matching | geodesic error: similar accuracy; ZoomOut doubles eigenfunction dimensions | No clear effect |
| use GSLR instead of ICP, BCICP, ZoomOut for shape matching | ICP, BCICP, ZoomOut · FAUST shape matching | coverage: covering more than 80% vertices on all the pairs | Improved |
Closest papers by meaning
- Shuffled Linear Regression with Outliers in Both Covariates and Responses (IJCV 2022 · similarity 0.92)
- Shuffled Deep Regression (AAAI 2024 · similarity 0.86)
- Distributed High-Dimensional Quantile Regression: Estimation Efficiency and Support Recovery (ICML 2024 · similarity 0.82)
- Agnostic Learning of Mixed Linear Regressions with EM and AM Algorithms (ICML 2024 · similarity 0.82)
- Linear Regression using Heterogeneous Data Batches (NeurIPS 2024 · similarity 0.82)
- Transfer Learning Meets Functional Linear Regression: No Negative Transfer Under Posterior Drift (AAAI 2025 · similarity 0.81)
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.
