Research map / Reinforcement learning / Policy gradient and actor-critic
Data-Efficient Policy Evaluation Through Behavior Policy Search
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 |
|---|---|---|---|
| Adapt the behavior policy with BPG rather than evaluate on-policy with Monte Carlo | Importance-sampling policy evaluation · 4x4 Gridworld, π1 | MSE: 85.48 % lower | Helped |
| Adapt the behavior policy with BPG rather than evaluate on-policy with Monte Carlo | Importance-sampling policy evaluation · 4x4 Gridworld, π2 | MSE: 31.02 % lower | Helped |
| Adapt the behavior policy with BPG rather than evaluate on-policy with Monte Carlo | Importance-sampling policy evaluation · Cart-pole Swing Up | MSE | Helped |
| Adapt the behavior policy with BPG rather than evaluate on-policy with Monte Carlo | Importance-sampling policy evaluation · Acrobot | MSE | Helped |
| Add behavior-policy search to the doubly robust estimator | DR-BPG · 10x10 stochastic Gridworld, fixed model | MSE: 56.9 % | Helped |
| Use a changing model rather than a fixed model for DR-BPG (ablation) | DR-BPG · 10x10 stochastic Gridworld | MSE | Mixed |
| Vary the BPG learning rate (ablation) | BPG · 4x4 Gridworld | IS variance | Mixed |
| Make the high-reward action less rare by increasing its probability (ablation) | BPG · Rare-event Gridworld | Relative variance reduction | Hurt |
| Use a fixed model instead of a continually updated model (ablation) | DR-BPG · 10x10 stochastic Gridworld | MSE | Helped |
Closest papers by meaning
- Efficient Policy Evaluation with Offline Data Informed Behavior Policy Design (ICML 2024 · similarity 0.90)
- SaVeR: Optimal Data Collection Strategy for Safe Policy Evaluation in Tabular MDP (ICML 2024 · similarity 0.90)
- Efficient Multi-Policy Evaluation for Reinforcement Learning (AAAI 2025 · similarity 0.89)
- Doubly Optimal Policy Evaluation for Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Efficient and Sharp Off-Policy Evaluation in Robust Markov Decision Processes (NeurIPS 2024 · similarity 0.88)
- Finding good policies in average-reward Markov Decision Processes without prior knowledge (NeurIPS 2024 · similarity 0.88)
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.