Research map / Reinforcement learning / Policy gradient and actor-critic
Continuous Deep Q-Learning with Model-based Acceleration
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 |
|---|---|---|---|
| NAF instead of DDPG | Q-learning on continuous control · Reacher | best test reward: -0.509 -> -0.331 | Helped |
| NAF instead of DDPG | Q-learning · Peg insertion | episodes to within 5% of best: 690 -> 130 | Helped |
| NAF instead of DDPG | Q-learning · Gripper | best test reward: 1.03 -> 1.81 | Helped |
| NAF instead of DDPG | Q-learning · GripperM | best test reward: -20.2 -> -12.4 | Helped |
| NAF instead of DDPG | Q-learning · Cheetah | best test reward: 8.23 -> 7.91 | Hurt |
| NAF instead of DDPG | Q-learning · Walker2d | best test reward: 2.96 -> 1.85 | Hurt |
| NAF instead of DDPG | Q-learning · Ant | best test reward: -2.54 -> -2.58 | No clear effect |
| mix iLQG/MPC off-policy episodes into replay (ablation) | NAF · Reacher, true dynamics | best test reward: -0.488 -> -0.449 | No clear effect |
| add imagination rollouts (ablation) | NAF · Reacher, true dynamics | episodes to best: 740 -> 450 | Helped |
| add imagination rollouts (ablation) | NAF · Canada2d, true dynamics | episodes to best: 1970 -> 570 | Helped |
| add imagination rollouts plus 0.5 MPC mix (ablation) | NAF · Canada2d, true dynamics | episodes to best: 1970 -> 140 | Helped |
| add imagination rollouts (ablation) | NAF · Cheetah, true dynamics | best test reward: 7.00 -> 7.36 | Helped |
Closest papers by meaning
- Actor-Free Continuous Control via Structurally Maximizable Q-Functions (NeurIPS 2025 · similarity 0.90)
- Learning from Suboptimal Data in Continuous Control via Auto-Regressive Soft Q-Network (ICML 2025 · similarity 0.90)
- Adaptive $Q$-Network: On-the-fly Target Selection for Deep Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities (NeurIPS 2026 · similarity 0.88)
- Continuous Q-Score Matching: Diffusion Guided Reinforcement Learning for Continuous-Time Control (NeurIPS 2025 · similarity 0.88)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · 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.