Research map / Reinforcement learning / Policy gradient and actor-critic
Importance Weighted Transfer of Samples in Reinforcement Learning
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 importance-weighted source samples, with separate reward and transition weights | IWFQI · Puddle World, shared dynamics | learning performance | Helped |
| Estimate importance weights with Gaussian processes instead of using ideal weights (ablation) | IWFQI · Puddle World, shared dynamics | learning performance | No clear effect |
| Use importance-weighted transfer instead of relevance-based transfer | IWFQI vs. RBT · Puddle World, shared dynamics | early learning performance | Helped |
| Account for task differences in transition dynamics when transferring samples | IWFQI · Puddle World, puddle-based dynamics | learning performance | Helped |
| Use relevance-based sample transfer instead of importance-weighted transfer | RBT vs. IWFQI · Puddle World, puddle-based dynamics | early learning performance | Hurt |
| Transfer samples under a shared-dynamics assumption when source and target dynamics differ | SDT · Puddle World, puddle-based dynamics | learning performance | Hurt |
| Use importance-weighted transfer instead of plain FQI | IWFQI · Acrobot swing-up | learning performance | Helped |
| Transfer reward and transition samples jointly rather than weighting them separately | RBT vs. IWFQI · Acrobot swing-up | transfer selectivity | Hurt |
| Use importance-weighted transfer when only the constant-spin source is available | IWFQI · Acrobot swing-up, constant-spin source only | learning performance | Helped |
| Transfer samples jointly when only the constant-spin source is available | RBT · Acrobot swing-up, constant-spin source only | learning performance | Hurt |
| Use importance-weighted transfer from existing reservoirs | IWFQI · Water Reservoir Control | average cost per day | Helped |
| Use relevance-based transfer instead of importance-weighted transfer | RBT vs. IWFQI · Water Reservoir Control | average cost per day | Hurt |
Closest papers by meaning
- Transfer of Samples in Policy Search via Multiple Importance Sampling (ICML 2019 · similarity 0.92)
- Prioritizing Samples in Reinforcement Learning with Reducible Loss (NeurIPS 2023 · similarity 0.89)
- Model-Based Transfer Learning for Contextual Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
- Pessimism Principle Can Be Effective: Towards a Framework for Zero-Shot Transfer Reinforcement Learning (ICML 2025 · similarity 0.85)
- Can RLHF be More Efficient with Imperfect Reward Models? A Policy Coverage Perspective (ICML 2025 · similarity 0.85)
- Reinforcement Learning Guided Semi-Supervised Learning (NeurIPS 2024 · 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.