Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-supervised learning for linear extremile regression
Rong Jiang, Jiangfeng Wang, Keming Yu
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 proposed SL linear extremile estimator instead of ordinary estimator | linear extremile regression · Sim 1, N(0,1), sigma=0.5, tau=0.1 | TAE: 0.526 -> 0.250 | Improved |
| use SL instead of OE at the median level | linear extremile regression · Sim 1, N(0,1), sigma=0.5, tau=0.5 | TAE: 0.179 -> 0.178 | No clear effect |
| use SL instead of OE with uniform errors | linear extremile regression · Sim 1, U(0,1), sigma=0.5, tau=0.1 | TAE: 0.203 -> 0.044 | Improved |
| use SL instead of OE with heavy-tailed errors | linear extremile regression · Sim 1, t(5), sigma=0.5, tau=0.9 | TAE: 0.808 -> 0.388 | Improved |
| use SL instead of OE under heteroscedastic noise | linear extremile regression · Sim 1, N(0,1), heteroscedastic, tau=0.9 | TAE: 0.513 -> 0.284 | Improved |
| add unlabeled data, increase N from 500 to 2000 (ablation) | SSL under model misspecification · Sim 2, N(0,1), tau=0.5, intercept | PARE (%): 71.6% -> 153.1% | Improved |
| add unlabeled data, increase N from 500 to 2000 (ablation) | SSL under model misspecification · Sim 2, U(0,1), tau=0.5, intercept | PARE (%): 85.8% -> 323.1% | Improved |
| add unlabeled data, increase N from 500 to 2000 (ablation) | SSL under model misspecification · Sim 2, t(5), tau=0.9, second coefficient | PARE (%): 4.2% -> 30.0% | Improved |
| use SSL with unlabeled data instead of SL | linear extremile regression · Homeless data, tau=0.9, MHI | PARE (%): 44.1% | Improved |
| use SSL instead of SL for the intercept | linear extremile regression · Homeless data, tau=0.1, Intercept | PARE (%): -11.4% | Hurt |
| use SSL instead of SL | linear extremile regression · Simulation 2, all settings | fraction of PARE above 20%: 82.2% | Improved |
Closest papers by meaning
- Online Quantile Regression (JMLR 2025 · similarity 0.86)
- Localized Debiased Machine Learning: Efficient Inference on Quantile Treatment Effects and Beyond (JMLR 2024 · similarity 0.85)
- Improved Generalization in Semi-Supervised Learning: A Survey of Theoretical Results (TPAMI 2022 · similarity 0.85)
- Semi-Supervised Regression with Heteroscedastic Pseudo-Labels (NeurIPS 2025 · similarity 0.85)
- Distributed High-Dimensional Quantile Regression: Estimation Efficiency and Support Recovery (ICML 2024 · similarity 0.85)
- Non-parametric Quantile Regression via the K-NN Fused Lasso (JMLR 2021 · similarity 0.84)
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.
