Research map / Reinforcement learning / Policy gradient and actor-critic
Trust Region Evolution Strategies
Guoqing Liu, Li Zhao, Feidiao Yang, Jiang Bian, Tao Qin, Nenghai Yu, Tie-Yan Liu
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 |
|---|---|---|---|
| add data reuse via clipped surrogate objective with multiple epochs (TRES) | ES · Swimmer-v1 | timesteps to threshold: 4.5e+5 -> 2.4e+5 | Improved |
| add clipped surrogate objective with multi-epoch reuse (TRES) | ES · HalfCheetah-v1 | timesteps to threshold: 1.7e+6 -> 9.8e+5 | Improved |
| add clipped surrogate objective with multi-epoch reuse (TRES) | ES · Walker2d-v1 | timesteps to threshold: 1.2e+7 -> 3.6e+6 | Improved |
| add clipped surrogate objective with multi-epoch reuse (TRES) | ES · Ant-v1 | timesteps to threshold: 1.5e+7 -> 1.3e+7 | Improved |
| add clipped surrogate objective with multi-epoch reuse (TRES) | ES · InvertedDoublePendulum-v1 | timesteps to threshold: 8.8e+5 -> 6e+5 | Improved |
| TRES vs PPO | PPO · Walker2d-v1 | timesteps to threshold: 1.9e+6 -> 3.6e+6 | Hurt |
| vary clip factor lambda away from 14.0 (too small or too large) (ablation) | TRES · HalfCheetah-v1 | training curve return | Hurt |
| clip factor lambda in [8.0, 20.0] vs baseline ES (ablation) | TRES · HalfCheetah-v1 | sample efficiency | Improved |
| increase epoch number K to large values (ablation) | TRES · HalfCheetah-v1 | training curve return | Hurt |
| use K=15 epochs rather than small K (ablation) | TRES · HalfCheetah-v1 | training curve return | Improved |
| multi-epoch updates in TRES vs single update in ES | TRES · HalfCheetah-v1 | wall-clock training time | No clear effect |
Closest papers by meaning
- Supported Trust Region Optimization for Offline Reinforcement Learning (ICML 2023 · similarity 0.89)
- Improving Exploration in Evolution Strategies for Deep Reinforcement Learning via a Population of Novelty-Seeking Agents (NeurIPS 2018 · similarity 0.88)
- Evolution-Guided Policy Gradient in Reinforcement Learning (NeurIPS 2018 · similarity 0.87)
- Signal-Adaptive Trust Regions for Gradient-Free Optimization of Recurrent Spiking Neural Networks (NeurIPS 2026 · similarity 0.85)
- Diffusion Policies Creating a Trust Region for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.85)
- Two-Stage Evolutionary Reinforcement Learning for Enhancing Exploration and Exploitation (AAAI 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.
