Research map / Learning theory and algorithms / Combinatorial and search optimization
Learning Optimal Predictive Checklists
Haoran Zhang, Quaid Morris, Berk Ustun, Marzyeh Ghassemi
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 |
|---|---|---|---|
| Fit checklists with MIP instead of unit weighting | Predictive checklist · 7 clinical datasets | training error: 7/7 datasets | Improved |
| Fit checklists with MIP instead of unit weighting | Predictive checklist · 7 clinical datasets | 5-fold CV test error: 5/7 datasets | Mixed |
| Relax M from 1 (MIP_OR) to general M (ablation) | MIP checklist · ptsd | test error: 7.8% | Improved |
| Relax M from 1 (MIP_OR) to general M (ablation) | MIP checklist · 7 clinical datasets | test error | Improved |
| Adaptive instead of fixed binarization (ablation) | MIP checklist · 7 clinical datasets | training error: 6/7 datasets | Improved |
| Adaptive instead of fixed binarization (ablation) | MIP checklist · 7 clinical datasets | test error: 5/7 datasets | Mixed |
| Adaptive instead of fixed binarization (ablation) | MIP checklist · adhd | test error: 17.1 -> 0.5 | Improved |
| Adaptive instead of fixed binarization (ablation) | MIP checklist · heart | test error: 28.8 -> 16.7 | Improved |
| Add group fairness constraints on FPR and FNR | MIP checklist · kidney (CRRT mortality) | test FPR gap Black Females vs White Males: 54.5% to 30.6% | Improved |
| Add group fairness constraints | MIP checklist · kidney (CRRT mortality) | overall FPR | Hurt |
| Use MIP checklist instead of complex models LR and XGB | Predictive checklist · mortality and kidney | test error | Hurt |
| Use checklist trained with MIP versus LR/XGB | Predictive checklist · heart and readmit | test error | No clear effect |
Closest papers by meaning
- Learning Optimized Risk Scores (JMLR 2019 · similarity 0.84)
- Learning Certifiably Optimal Rule Lists for Categorical Data (JMLR 2018 · similarity 0.83)
- Optimal Decision-Making Based on Prediction Sets (ICML 2026 · similarity 0.82)
- Aligning Evaluation with Clinical Priorities: Calibration, Label Shift, and Error Costs (NeurIPS 2025 · similarity 0.82)
- Overcoming Common Flaws in the Evaluation of Selective Classification Systems (NeurIPS 2024 · similarity 0.80)
- Decision Theoretic Foundations for Conformal Prediction: Optimal Uncertainty Quantification for Risk-Averse Agents (ICML 2025 · similarity 0.80)
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.
