Research map / Reinforcement learning / Policy gradient and actor-critic
Reliable Off-Policy Learning for Dosage Combinations
Jonas Schweisthal, Dennis Frauen, Valentyn Melnychuk, Stefan Feuerriegel
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 reliable policy selection (overlap constraint) in place of naive selection (ablation) | DCNet · TCGA | regret (selected policy): 2.52 -> 0.03 | Improved |
| add reliable policy selection in place of naive selection (ablation) | DCNet · MIMIC-IV | regret (selected policy): 2.81 -> 0.04 | Improved |
| add reliable policy selection in place of naive selection (ablation) | VCNet · TCGA | regret (selected policy): 3.64 -> 0.06 | Improved |
| add reliable policy selection in place of naive selection (ablation) | VCNet · MIMIC-IV | regret (selected policy): 2.13 -> 0.07 | Improved |
| add reliable policy selection in place of naive selection (ablation) | MLP · TCGA | regret (selected policy): 1.94 -> 1.78 | Improved |
| add reliable policy selection in place of naive selection (ablation) | MLP · MIMIC-IV | regret (selected policy): 0.02 -> 0.03 | No clear effect |
| replace MLP with DCNet dose-response estimator (with reliable selection) (ablation) | reliable policy learning · TCGA | regret (selected policy): 1.78 -> 0.03 | Improved |
| replace MLP with DCNet dose-response estimator (with reliable selection) (ablation) | reliable policy learning · MIMIC-IV | regret (selected policy): 0.03 -> 0.04 | No clear effect |
| replace VCNet with DCNet (modeling joint effect of dosages) with reliable selection (ablation) | reliable policy learning · TCGA | regret (selected policy): 0.06 -> 0.03 | Improved |
| replace VCNet with DCNet with reliable selection (ablation) | reliable policy learning · MIMIC-IV | regret (selected policy): 0.07 -> 0.04 | Improved |
| reliable selection reduces std of regret across restarts (ablation) | DCNet · TCGA | regret std: 1.21 -> 0.00 | Improved |
| discretize dosages (5x5 grid, oracle mu) instead of continuous dosages (ablation) | oracle policy · MIMIC-IV | regret: 0.04 -> 0.05 | Hurt |
Closest papers by meaning
- Contrastive Balancing Representation Learning for Heterogeneous Dose-Response Curves Estimation (AAAI 2024 · similarity 0.86)
- Enhancing Treatment Effect Estimation via Active Learning: A Counterfactual Covering Perspective (ICML 2025 · similarity 0.85)
- Doubly Robust Fusion of Many Treatments for Policy Learning (ICML 2025 · similarity 0.85)
- Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data (NeurIPS 2025 · similarity 0.84)
- PairNet: Training with Observed Pairs to Estimate Individual Treatment Effect (ICML 2024 · similarity 0.84)
- Quantifying Aleatoric Uncertainty of the Treatment Effect: A Novel Orthogonal Learner (NeurIPS 2024 · similarity 0.84)
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.
