Research map / Reinforcement learning / Offline RL
Semi-Supervised Off-Policy Reinforcement Learning and Value Estimation for Dynamic Treatment Regimes
Aaron Sonabend-W, Nilanjana Laha, Ashwin N. Ananthakrishnan, Tianxi Cai, Rajarshi Mukherjee
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 unlabeled data via semi-supervised imputation | supervised Q-learning · Simulation gamma11 (n=500, N=10000) | empirical standard error: 0.22 -> 0.12 | Improved |
| add unlabeled data via semi-supervised imputation | supervised Q-learning · Simulation gamma12 (n=500, N=10000) | empirical standard error: 0.29 -> 0.17 | Improved |
| use basis expansion instead of random forest imputation | SSL Q-learning · Simulation gamma21 (n=500, N=10000) | relative efficiency: 1.80 -> 1.86 | Improved |
| add unlabeled data to the doubly robust value estimator | supervised DR value estimator · EHR setting, correct models | relative efficiency: 1.55 | Improved |
| add unlabeled data to the doubly robust value estimator | supervised DR value estimator · EHR setting, misspecified Q | empirical standard error: 0.07 -> 0.04 | Improved |
| add unlabeled data to the doubly robust value estimator | supervised DR value estimator · Continuous setting, misspecified propensity | empirical standard error: 0.14 -> 0.11 | Improved |
| add unlabeled data to the value estimator | supervised DR value estimator · IBD data | standard error: 0.486 -> 0.397 | Improved |
| add unlabeled data to the value estimator | supervised DR value estimator · IBD data | value estimate: 0.851 -> 0.871 | Improved |
| misspecify the Q function model | SSL and supervised estimators · Simulation interaction parameters | bias | No clear effect |
| use semi-supervised Q-learning | supervised Q-learning · IBD data | power to detect significant predictors | Improved |
Closest papers by meaning
- EpiCare: A Reinforcement Learning Benchmark for Dynamic Treatment Regimes (NeurIPS 2024 · similarity 0.90)
- Position: Reinforcement Learning in Dynamic Treatment Regimes Needs Critical Reexamination (ICML 2024 · similarity 0.89)
- Offline Guarded Safe Reinforcement Learning for Medical Treatment Optimization Strategies (NeurIPS 2025 · similarity 0.87)
- Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data (NeurIPS 2025 · similarity 0.85)
- Towards Safe Policy Learning under Partial Identifiability: A Causal Approach (AAAI 2024 · similarity 0.85)
- Enhancing Treatment Effect Estimation via Active Learning: A Counterfactual Covering Perspective (ICML 2025 · 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.
