Research map / Reinforcement learning / Policy gradient and actor-critic
Generalized Proximal Policy Optimization with Sample Reuse
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 |
|---|---|---|---|
| Reuse samples from recent policies with generalized clipping and adaptive learning rate (GePPO instead of PPO) | PPO · Swimmer-v3 | average performance over 1M steps: 98 -> 161 | Helped |
| Reuse samples from recent policies with generalized clipping and adaptive learning rate (GePPO instead of PPO) | PPO · Hopper-v3 | average performance over 1M steps: 2{,}362 -> 2{,}544 | Helped |
| Reuse samples from recent policies with generalized clipping and adaptive learning rate (GePPO instead of PPO) | PPO · HalfCheetah-v3 | average performance over 1M steps: 1{,}764 -> 2{,}439 | Helped |
| Reuse samples from recent policies with generalized clipping and adaptive learning rate (GePPO instead of PPO) | PPO · Walker2d-v3 | average performance over 1M steps: 1{,}817 -> 2{,}199 | Helped |
| Reuse samples from recent policies with generalized clipping and adaptive learning rate (GePPO instead of PPO) | PPO · Ant-v3 | average performance over 1M steps: 545 -> 762 | Helped |
| Reuse samples from recent policies with generalized clipping and adaptive learning rate (GePPO instead of PPO) | PPO · Humanoid-v3 | average performance over 1M steps: 584 -> 665 | Helped |
| Use GePPO instead of PPO | PPO · Swimmer-v3 | steps (M) to final PPO performance: 1.00 -> 0.23 | Helped |
| Use GePPO instead of PPO | PPO · Hopper-v3 | steps (M) to final PPO performance: 1.00 -> 0.41 | Helped |
| Use GePPO instead of PPO | PPO · HalfCheetah-v3 | steps (M) to final PPO performance: 1.00 -> 0.54 | Helped |
| Use GePPO instead of PPO | PPO · Walker2d-v3 | steps (M) to final PPO performance: 1.00 -> 0.63 | Helped |
| Use GePPO instead of PPO | PPO · Ant-v3 | steps (M) to final PPO performance: 1.00 -> 0.79 | Helped |
| Use GePPO instead of PPO | PPO · Humanoid-v3 | steps (M) to final PPO performance: 1.00 -> 0.85 | Helped |
Closest papers by meaning
- Off-Policy Proximal Policy Optimization (AAAI 2023 · similarity 0.91)
- Robust On-Policy Sampling for Data-Efficient Policy Evaluation in Reinforcement Learning (NeurIPS 2022 · similarity 0.91)
- Iteratively Refined Behavior Regularization for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Sample-Efficiency in Multi-Batch Reinforcement Learning: The Need for Dimension-Dependent Adaptivity (ICLR 2024 · similarity 0.89)
- Enhancing Efficiency of Safe Reinforcement Learning via Sample Manipulation (NeurIPS 2024 · similarity 0.89)
- Doubly Optimal Policy Evaluation for Reinforcement Learning (ICLR 2025 · 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.