Research map / Label-efficient and robust learning / Semi-supervised learning
Prior Adaptive Semi-supervised Learning with Application to EHR Phenotyping
Yichi Zhang, Molei Liu, Matey Neykov, Tianxi Cai
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 |
|---|---|---|---|
| add prior-adaptive shrinkage toward the surrogate direction (PASS) versus supervised LASSO with equal labels | logistic LASSO · Simulation scenario I (n=100 labels) | AUC, excess risk, MSE-P: PASS_100 outperforms LASSO_400 | Improved |
| add PASS to supervised LASSO | logistic LASSO · Simulation scenarios II-V (n=100 labels) | AUC, excess risk, MSE-P: PASS_100 similar to LASSO_200, better than LASSO_150 | Improved |
| use PASS when surrogate is unreliable instead of LASSO | logistic LASSO · Simulation scenario VI (n=100 labels) | AUC, excess risk, MSE-P: comparable to LASSO_100 | No clear effect |
| use PASS instead of SS-prior and pLASSO when directions differ | SS-prior, pLASSO · Simulation scenario II (n=100 labels) | AUC, excess risk, MSE-P | Improved |
| use SS-prior instead of PASS when prior holds exactly | PASS · Simulation scenario I (n=100 labels) | AUC, excess risk, MSE-P | No clear effect |
| use adaptive LASSO weights instead of LASSO | supervised logistic LASSO · Simulation scenarios I-VI (n=100 labels) | AUC, excess risk, MSE-P | Hurt |
| add PASS to supervised estimators under surrogate misspecification | supervised LASSO · Misspecified simulation scenarios i-iii | AUC, excess risk, MSE-P: outperforms supervised LASSO using 2-3 more times of sample sizes | Improved |
| violate the conditional independence prior assumption (scenario ii) for SS-prior and pLASSO2 | SS-prior, pLASSO2 · Misspecified simulation scenario ii | AUC, excess risk, MSE-P | Hurt |
| use PASS instead of pLASSO1 with denser eta | pLASSO1 · Misspecified simulation scenario iii | AUC, excess risk, MSE-P | Improved |
| add PASS to supervised LASSO/ALASSO at small label size | supervised LASSO and ALASSO · EHR phenotyping (CAD, RA, Depression), n=50 | AUC, Brier skill score | Improved |
| increase label size n | PASS versus supervised · EHR phenotyping (CAD, RA, Depression) | AUC, Brier skill score | No clear effect |
| use PASS instead of SS-prior and SS-ULASSO | SS-prior, SS-ULASSO · Depression phenotyping EHR (n=50 labels) | AUC: about 0.1 higher AUC; 0.05 higher than pLASSO | Improved |
Closest papers by meaning
- Surrogate Assisted Semi-supervised Inference for High Dimensional Risk Prediction (JMLR 2023 · similarity 0.90)
- SMART: Towards Pre-trained Missing-Aware Model for Patient Health Status Prediction (NeurIPS 2024 · similarity 0.84)
- Systematic Comparison of Semi-supervised and Self-supervised Learning for Medical Image Classification (CVPR 2024 · similarity 0.83)
- Time-to-Event Pretraining for 3D Medical Imaging (ICLR 2025 · similarity 0.83)
- ACES: Automatic Cohort Extraction System for Event-Stream Datasets (ICLR 2025 · similarity 0.83)
- One Loss to Rule Them All: Marked Time-to-Event for Structured EHR Foundation Models (NeurIPS 2026 · similarity 0.83)
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.
