Research map / Reinforcement learning / Policy gradient and actor-critic
Episodic Policy Gradient Training
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 |
|---|---|---|---|
| Use EPGT instead of the strongest listed baseline (TMG) | A2C · HalfCheetah | mean return: 1,568 -> 2,530 | Helped |
| Use EPGT instead of the strongest listed baseline (HOOF, our run) | A2C · Hopper | mean return: 452 ± 40.7 -> 603 ± 187 | Helped |
| Use EPGT instead of the strongest listed baseline (HOOF, our run) | A2C · Ant | mean return: 954 ± 8.57 -> 1,083 ± 126 | Helped |
| Use EPGT instead of the strongest listed baseline (HOOF, our run) | A2C · Walker | mean return: 674 ± 195 -> 888 ± 425 | Helped |
| Use EPGT instead of PB2 (our run) | PPO · BW | best median score: 280 -> 282 | Helped |
| Use EPGT instead of PB2 (reported) | PPO · LLC | best median score: 235 -> 235 | No clear effect |
| Use EPGT instead of PB2 (reported) | PPO · Hopper | best median score: 2,346 -> 3,253 | Helped |
| Use EPGT instead of PB2 (our run) | PPO · IDP | best median score: 9,253 -> 9,322 | Helped |
| Learn the hyper-state projection and embedding with a VAE rather than random mapping (ablation) | EPGT with A2C · Mountain Car | return | Helped |
| Increase writing-neighbor count from 1 to 3 (ablation) | EPGT with PPO · Alien | score: around 500 score | Helped |
| Increase writing-neighbor count beyond 3 (ablation) | EPGT with PPO · Alien | score | No clear effect |
| Increase representation order from 0 to 2, adding derivative information (ablation) | EPGT with PPO · MsPacman | return | Helped |
Closest papers by meaning
- e-COP : Episodic Constrained Optimization of Policies (NeurIPS 2024 · similarity 0.89)
- Model-Based Episodic Memory Induces Dynamic Hybrid Controls (NeurIPS 2021 · similarity 0.89)
- Learning Self-Imitating Diverse Policies (ICLR 2019 · similarity 0.88)
- Open the Black Box: Step-based Policy Updates for Temporally-Correlated Episodic Reinforcement Learning (ICLR 2024 · similarity 0.88)
- Efficient Episodic Memory Utilization of Cooperative Multi-Agent Reinforcement Learning (ICLR 2024 · similarity 0.87)
- Discovering Temporally-Aware Reinforcement Learning Algorithms (ICLR 2024 · similarity 0.87)
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.