Research map / Reinforcement learning / MDPs and planning
Practical Kernel-Based Reinforcement Learning
André M.S. Barreto, Doina Precup, Joelle Pineau
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 |
|---|---|---|---|
| replace KBRL(8000) with KBSF(8000,100) (compress model to 100 representative states) | KBRL · Puddle world | operations to find a policy: decrease of approximately 99.58% | Improved |
| increase number of representative states m toward n (ablation) | KBSF(8000,m) · Puddle world | return | Improved |
| use KBSF instead of LSPI with same data and architecture | LSPI · Single pole-balancing | success rate: no more than 60% -> 90% | Improved |
| use KBSF(10^6,200) on double pole-balancing | KBSF · Double pole-balancing | success rate: more than 80% | Improved |
| increase number of representative states m in KBSF (ablation) | KBSF · HIV drug schedule | return | No clear effect |
| decrease eta_min in FQIT (ablation) | FQIT · HIV drug schedule | return | Improved |
| add a committee of independently trained KBSF agents with majority voting (ablation) | KBSF · HIV drug schedule | return: 5 agents comparable to FQIT(50); 15 agents exceed best FQIT | Improved |
| use KBSF(10000) instead of FQIT for run time | FQIT · HIV drug schedule | run time: 15 times faster than FQTI(100), 20 times faster than FQTI(50) | Improved |
| use KBSF instead of LSPI or FQIT(200) with varied seizure penalty | LSPI / FQIT(200) · Epilepsy suppression | seizure vs stimulation tradeoff: at least 100 times faster than the tree-based method | Improved |
Closest papers by meaning
- Reward-Free Kernel-Based Reinforcement Learning (ICML 2024 · similarity 0.91)
- Kernel-Based Function Approximation for Average Reward Reinforcement Learning: An Optimist No-Regret Algorithm (NeurIPS 2024 · similarity 0.90)
- Reinforcement Learning in Feature Space: Matrix Bandit, Kernels, and Regret Bound (ICML 2020 · similarity 0.90)
- Pass@K Policy Optimization: Solving Harder Reinforcement Learning Problems (NeurIPS 2025 · similarity 0.88)
- Kernel Metric Learning for In-Sample Off-Policy Evaluation of Deterministic RL Policies (ICLR 2024 · similarity 0.87)
- Iteratively Refined Behavior Regularization for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
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.
