Research map / Reinforcement learning / Visual and representation RL
Sample-Efficient Automated Deep 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 |
|---|---|---|---|
| Share collected replay experience across the population (ablation) | SEARL optimizing TD3 · MuJoCo | performance | Helped |
| Add architecture adaptation (ablation) | SEARL optimizing TD3 · HalfCheetah | performance | Helped |
| Add architecture adaptation (ablation) | SEARL optimizing TD3 · Walker2D | performance | Hurt |
| Add hyperparameter adaptation (ablation) | SEARL optimizing TD3 · MuJoCo | performance | Mixed |
| Re-create the target network and re-initialize the optimizer after each episode (ablation) | TD3 · MuJoCo | performance | No clear effect |
| Randomly initialize the population's hyperparameters instead of using the default initialization (ablation) | SEARL · MuJoCo | performance | No clear effect |
| Use shared replay and online population-based optimization instead of random search or modified PBT | TD3 · MuJoCo | environment interactions to reach the same reward: up to $10\times$ | Helped |
| Use SEARL rather than random search | Rainbow DQN · Atari | environment interactions to reach the same reward: an order of magnitude | Helped |
| Use SEARL rather than random search | Rainbow DQN · Freeway | final performance | Hurt |
| Use SEARL rather than random search | Rainbow DQN · Pong and RoadRunner | final performance | No clear effect |
| Use SEARL rather than random search | Rainbow DQN · Enduro and Boxing | final performance | Helped |
| Optimize architecture and hyperparameters online rather than use a held-out tuned TD3 configuration | SEARL · Held-out MuJoCo environments | performance | Mixed |
Closest papers by meaning
- Adaptive $Q$-Network: On-the-fly Target Selection for Deep Reinforcement Learning (ICLR 2025 · similarity 0.91)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.90)
- ESRL: Efficient Sampling-Based Reinforcement Learning for Sequence Generation (AAAI 2024 · similarity 0.89)
- FASTER: Value-Guided Sampling for Fast RL (NeurIPS 2026 · similarity 0.89)
- Parallelizing Model-based Reinforcement Learning Over the Sequence Length (NeurIPS 2024 · similarity 0.89)
- EfficientZero V2: Mastering Discrete and Continuous Control with Limited Data (ICML 2024 · similarity 0.89)
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.