Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Learning with Noisy Proxy Covariates: Generalization Bounds and Distribution Regression
Kwangho Kim, Jisu Kim
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 all proxy covariates for feature learning instead of labeled only | SSDR · Synthetic Euclidean, tau=.10, N=1000 | normalized RMSE: .113 -> .057 | Improved |
| increase total proxy count N from 100 to 1000 | SSDR · Synthetic Euclidean, tau=.10 | normalized RMSE: .091 -> .057 | Improved |
| increase total proxy count N from 100 to 1000 | SSDR · Synthetic Euclidean, tau=.40 | normalized RMSE: .176 -> .142 | Improved |
| increase proxy noise tau from .10 to .40 | SSDR · Synthetic Euclidean, N=1000 | normalized RMSE: .057 -> .142 | Hurt |
| add more unlabeled proxies on label-only baseline | Label-only spectral · Synthetic Euclidean, tau=.10 | normalized RMSE: .112 -> .113 | No clear effect |
| use SSDR instead of LapRLS at low label fraction | SSDR · Default risk FnGuide, 5% labeled | normalized RMSE: .07 -> .06 | Improved |
| use SSDR instead of LapRLS at high label fraction | SSDR · Default risk FnGuide, 40% labeled | normalized RMSE: .04 -> .05 | Hurt |
| use SSDR instead of best supervised baseline | SSDR · Default risk Bloomberg, 40% labeled | normalized RMSE: .07 -> .06 | Improved |
Closest papers by meaning
- Towards Self-Supervised Covariance Estimation in Deep Heteroscedastic Regression (ICLR 2025 · similarity 0.87)
- Leveraging Data Symmetries to Select an Optimal Subset of Training Data under Label Noise (NeurIPS 2026 · similarity 0.86)
- Semi-Supervised Regression with Heteroscedastic Pseudo-Labels (NeurIPS 2025 · similarity 0.86)
- SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning (ICML 2024 · similarity 0.85)
- Out-Of-Domain Unlabeled Data Improves Generalization (ICLR 2024 · similarity 0.85)
- Label Noise: Ignorance Is Bliss (NeurIPS 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.
