Research map / Learning theory and algorithms / Matrix and tensor methods
Robust Geodesic Regression
Ha-Young Shin, Hee‐Seok Oh
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 robust L1 estimator instead of L2 least squares | geodesic regression · Corpus callosum, tampered (20 of 88 flipped) | distance of p-hat to baseline: 0.1444551 -> 0.0182806 | Improved |
| use Tukey biweight instead of L2 | geodesic regression · Corpus callosum, tampered | distance of p-hat to baseline: 0.1444551 -> 0.0129771 | Improved |
| use L1 instead of Tukey biweight on clean data (ablation) | geodesic regression · Corpus callosum, untampered | distance of p-hat to baseline: 0.0061325 -> 0.0018924 | Improved |
| use Tukey instead of L1 on tampered data (ablation) | geodesic regression · Corpus callosum, tampered | distance of p-hat to baseline: 0.0182806 -> 0.0129771 | Improved |
| use L1 estimator instead of L2 under heavy-tailed errors | geodesic regression · S^2/H^2/S^3/H^3 simulation, type T and C noise | MSE | Improved |
| use L2 instead of robust estimators under normal errors | geodesic regression · S^n and H^n simulation, type N errors | MSE | Improved |
| use L1 estimator at very small sample size | geodesic regression · simulations, N=4 or 8 | MSE | Improved |
| use L1 vs L2 as dimension grows | L1 estimator · normal errors, n=10 | asymptotic relative efficiency: 0.84883 -> 0.95131 | Improved |
| increase curvature scale (sigma pi/4 to pi/2) (ablation) | Huber and Tukey estimators · S^3 normal errors | sample relative efficiency to L2: 0.9637054 -> 1.0052173 | Improved |
| increase spread sigma from pi/32 to pi/4 on S^3 (ablation) | L1 estimator · S^3 normal errors | sample relative efficiency to L2: 0.8408682 -> 0.8490783 | No clear effect |
Closest papers by meaning
- Riemannian Ordinary Least Squares (NeurIPS 2026 · similarity 0.90)
- Differentially Private Geodesic Regression (ICML 2026 · similarity 0.89)
- Multivariate Regression with Gross Errors on Manifold-Valued Data (TPAMI 2018 · similarity 0.89)
- RMLR: Extending Multinomial Logistic Regression into General Geometries (NeurIPS 2024 · similarity 0.87)
- Intrinsic Riemannian Cross-covariance for Manifold-valued Random Objects (NeurIPS 2026 · similarity 0.86)
- Non-parametric Online Change Point Detection on Riemannian Manifolds (ICML 2024 · 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.
