Research map / Learning theory and algorithms / Prediction and decision losses
Learning Optimal Dynamic Treatment Regimens Subject to Stagewise Risk Controls
Mochuan Liu, Yuanjia Wang, Haoda Fu, Donglin Zeng
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 |
|---|---|---|---|
| compare BR-DTRs with AOWL that ignores risk constraints | BR-DTRs · Simulation Setting I (linear kernel, n=400) | stage I risk: 2.678 -> 1.400 | Improved |
| compare BR-DTRs with AOWL that ignores risk constraints | BR-DTRs · Simulation Setting II (Gaussian kernel, n=400) | cumulative reward: 5.188 -> 3.069 | Hurt |
| add delayed-effect adjustment (BR-DTRs instead of naive Y1+Y2 outcome) (ablation) | BR-DTRs · Simulation Setting II (linear kernel, n=400) | cumulative reward: 1.797 -> 3.174 | Improved |
| add delayed-effect adjustment (BR-DTRs instead of naive Y1+Y2 outcome) (ablation) | BR-DTRs · Simulation Setting I (Gaussian kernel, n=400) | cumulative reward: 2.201 -> 2.301 | Improved |
| use Gaussian kernel instead of linear kernel under nonlinear boundaries | BR-DTRs · Simulation Setting II (n=400, eta=0.1) | cumulative reward: 3.169 -> 3.009 | Hurt |
| increase shifting parameter eta from 0.02 to 0.1 with Gaussian kernel (ablation) | BR-DTRs · Simulation Setting II (Gaussian kernel, n=400) | cumulative reward: 3.069 -> 3.009 | Hurt |
| increase shifting parameter eta from 0.02 to 0.1 with linear kernel (ablation) | BR-DTRs · Simulation Setting II (linear kernel, n=400) | Reward - II: 1.362 -> 1.177 | Hurt |
| increase training sample size from 200 to 400 (ablation) | BR-DTRs · Simulation Setting I | reward/risk accuracy | Improved |
| increase training sample size from 200 to 400 with linear kernel (ablation) | BR-DTRs · Simulation Setting II | reward/risk accuracy | No clear effect |
| tighten stage II risk constraint tau2 from infinity to 0.334 (tau1=1.005) | BR-DTRs · DURABLE T2D study | reward: 1.620 -> 1.547 | Hurt |
| tighten stage I risk constraint tau1 from infinity to 1.005 (tau2=infinity) | BR-DTRs · DURABLE T2D study | reward: 1.713 -> 1.620 | Hurt |
| BR-DTRs instead of naive under tau2=0.334, tau1=0.948 (ablation) | BR-DTRs · DURABLE T2D study | reward: 1.499 -> 1.520 | Improved |
Closest papers by meaning
- Gradient Regularized V-Learning for Dynamic Treatment Regimes (NeurIPS 2020 · similarity 0.90)
- Doubly Robust Fusion of Many Treatments for Policy Learning (ICML 2025 · similarity 0.85)
- Evaluating and Learning Optimal Dynamic Treatment Regimes under Truncation by Death (NeurIPS 2025 · similarity 0.85)
- Position: Reinforcement Learning in Dynamic Treatment Regimes Needs Critical Reexamination (ICML 2024 · similarity 0.84)
- Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data (NeurIPS 2025 · similarity 0.84)
- SAFER: A Calibrated Risk-Aware Multimodal Recommendation Model for Dynamic Treatment Regimes (ICML 2025 · 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.
