Research map / Reinforcement learning / Policy gradient and actor-critic
An Efficient Asynchronous Method for Integrating Evolutionary and Gradient-based Policy Search
Kyunghyun Lee, Byeong-Uk Lee, Ukcheol Shin, In So Kweon
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 synchronous CEM-RL with AES-RL | CEM-RL · HalfCheetah-v2 | mean score: 10725 -> 12550 | Improved |
| replace synchronous CEM-RL with AES-RL | CEM-RL · Hopper-v2 | mean score: 3613 -> 3751 | Improved |
| replace synchronous CEM-RL with AES-RL | CEM-RL · Walker2D-v2 | mean score: 4711 -> 5474 | Improved |
| replace synchronous CEM-RL with AES-RL | CEM-RL · Ant-v2 | mean score: 4251 -> 5120; SAC 5985 | Mixed |
| replace synchronous CEM-RL with AES-RL | CEM-RL · Swimmer-v2 | mean score: 75 -> 161 | Improved |
| replace synchronous CEM-RL with AES-RL | CEM-RL · Humanoid-v2 | mean score: 5579 -> 6136 | Improved |
| use five-worker parallel asynchronous execution instead of serial synchronous execution | CEM-RL · HalfCheetah-v2 | training time: 467 -> 83 min | Improved |
| use five-worker parallel asynchronous execution instead of serial synchronous execution | CEM-RL · Walker2d-v2 | training time: 487 -> 105 min | Improved |
| use five-worker parallel asynchronous execution instead of serial synchronous execution | CEM-RL · Hopper-v2 | training time: 505 -> 133 min | Improved |
| add asynchronous updates without the proposed mean and variance update rules (ablation) | CEM-RL · HalfCheetah-v2 | mean score: 10725 -> 10010 | Hurt |
| add asynchronous updates without the proposed mean and variance update rules (ablation) | CEM-RL · Walker2D-v2 | mean score: 4711 -> 4230 | Hurt |
| use the proposed variance update instead of fixing variance at 0.0001 (ablation) | AES-RL · HalfCheetah-v2 | mean score: 11636 -> 12550 | Improved |
Closest papers by meaning
- ERL-TD: Evolutionary Reinforcement Learning Enhanced with Truncated Variance and Distillation Mutation (AAAI 2024 · similarity 0.88)
- Asynchronous Federated Reinforcement Learning with Policy Gradient Updates: Algorithm Design and Convergence Analysis (ICLR 2025 · similarity 0.87)
- Value-Evolutionary-Based Reinforcement Learning (ICML 2024 · similarity 0.87)
- The Ladder in Chaos: Improving Policy Learning by Harnessing the Parameter Evolving Path in A Low-dimensional Space (NeurIPS 2024 · similarity 0.87)
- EvoRainbow: Combining Improvements in Evolutionary Reinforcement Learning for Policy Search (ICML 2024 · similarity 0.86)
- Learning to Reuse Policies in State Evolvable Environments (ICML 2025 · 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.
