Research map / Reinforcement learning / Visual and representation RL
Neural Episodic Control with State Abstraction
Zhuo Li, Derui Zhu, Yujing Hu, Xiaofei Xie, Lei Ma, YAN ZHENG, Yan Song, Yingfeng Chen, Jianjun Zhao
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 |
|---|---|---|---|
| use three-step rather than one-step or two-step abstract patterns (ablation) | NECSA · Walker2d-v3 and Hopper-v3 | sample efficiency | Improved |
| use longer patterns beyond three steps (ablation) | NECSA · MuJoCo tasks | sample efficiency | Hurt |
| measure patterns with reward confidence scores instead of Q-values (ablation) | NECSA · MuJoCo evaluation | sample efficiency | Improved |
| abstract state-action pairs instead of states alone (ablation) | NECSA · MuJoCo tasks | performance | Improved |
| abstract policy hidden outputs instead of raw image states (ablation) | NECSA · Atari games | performance | Improved |
| increase grid granularity (ablation) | NECSA · Walker2d-v3, Hopper-v3, and Alien | performance | Mixed |
| set reward-shaping magnitude epsilon within the reported best range (ablation) | NECSA · Walker2d-v3 and Hopper-v3 | performance | Improved |
| set epsilon greater or smaller than the reported best range (ablation) | NECSA · Walker2d-v3 and Hopper-v3 | performance | Hurt |
| add NECSA to the backbone reinforcement-learning algorithm (ablation) | TD3 and DDPG · Walker2d-v3 and Hopper-v3 | sample efficiency | Improved |
| add NECSA to DrQ-v2 (ablation) | DrQ-v2 · DMControl Walker-Walk and Hopper-Stand | sample efficiency | Improved |
| add NECSA to DrQ-v2 (ablation) | DrQ-v2 · DMControl Cartpole-Balance | performance | No clear effect |
| use Gaussian random projection to reduce state or hidden-output dimensions (ablation) | NECSA · high-dimensional MuJoCo and image-state tasks | performance | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Rainbow — Rainbow: Combining Improvements in Deep Reinforcement Learning (AAAI 2018) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Open the Black Box: Step-based Policy Updates for Temporally-Correlated Episodic Reinforcement Learning (ICLR 2024 · similarity 0.85)
- DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization (ICLR 2024 · similarity 0.84)
- e-COP : Episodic Constrained Optimization of Policies (NeurIPS 2024 · similarity 0.84)
- Efficient Episodic Memory Utilization of Cooperative Multi-Agent Reinforcement Learning (ICLR 2024 · similarity 0.84)
- TOP-ERL: Transformer-based Off-Policy Episodic Reinforcement Learning (ICLR 2025 · similarity 0.83)
- RVI-SAC: Average Reward Off-Policy Deep Reinforcement Learning (ICML 2024 · similarity 0.83)
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.
