Research map / Reinforcement learning / Policy gradient and actor-critic
Efficient Policy Learning from Surrogate-Loss Classification Reductions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use DeepGMM instead of empirical surrogate-loss minimization | Policy learning · Synthetic scenarios; linear and flexible policy classes | relative mean regret reduction (RMRR): typically with around 10% to 20% RMRR | Mixed |
| Use DeepGMM instead of EntropyLearning | Linear policy · Jobs case study | predicted policy value (×1000): -0.96 -> 4.42 | Helped |
| Use DeepGMM instead of EntropyLearning | Flexible policy · Jobs case study | predicted policy value (×1000): -1.75 -> 7.68 | Helped |
| Use DeepGMM instead of EntropyLearning | Linear policy with correctly specified model · Linear synthetic scenarios | parameter squared error | Helped |
| Fit nuisance functions with flexible neural networks instead of correctly specified linear or logistic regression (ablation) | DeepGMM and EntropyLearning · Quadratic synthetic scenarios | relative mean regret reduction | No clear effect |
| Use finite GMM with polynomial or random-feature critics instead of DeepGMM | Policy learning · Synthetic scenarios | relative mean regret reduction | Hurt |
Closest papers by meaning
- Off-policy Evaluation Beyond Overlap: Sharp Partial Identification Under Smoothness (ICML 2024 · similarity 0.85)
- Decision-Focused Learning with Directional Gradients (NeurIPS 2024 · similarity 0.84)
- Distributionally Robust Policy Learning under Concept Drifts (ICML 2025 · similarity 0.83)
- One-Shot Strategic Classification Under Unknown Costs (ICML 2024 · similarity 0.83)
- Policy Learning for Balancing Short-Term and Long-Term Rewards (ICML 2024 · similarity 0.83)
- Q-MMR: Off-Policy Evaluation via Recursive Reweighting and Moment Matching (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.