Research map / Reinforcement learning / Policy gradient and actor-critic
On the Computational Limits of AI4S-RL : A Unified $\varepsilon$-$N$ Analysis
Qili Shen, Kairui Feng, Ang He, Xuanhong Chen, Dake Zhang
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 |
|---|---|---|---|
| move resolution parameters away from the optimum (K=1.5, log ratio 1/3) in tabular RL (ablation) | tabular RL on Cart-Pole · Cart-Pole sample complexity | samples to 1% value error: 4500 -> 46000 | Hurt |
| off-optimal resolution choices instead of optimal ones (ablation) | tabular RL and Q-learning on Cart-Pole · Cart-Pole sample complexity | computational cost: N^1.6 | Hurt |
| use Q-learning with function approximation instead of tabular RL | Cart-Pole resolution study · Cart-Pole sample complexity | samples to 1% value error: 4500 -> 450 | Improved |
| increase surrogate resolution monotonically (ablation) | tabular RL on Cart-Pole · Cart-Pole sample complexity | log10(N): 3.65 -> 4.66 | Mixed |
| coarsen spatial ratio to K=8.0 (ablation) | PPO on Teppanyaki heat sequencing · Teppanyaki PDE control | convergence: K = 8.0 fail to converge | Hurt |
| choose theory-predicted K*=6 and log ratio 1/3 | PPO on Teppanyaki heat sequencing · Teppanyaki PDE control | total computational cost: K*=6, loghw(hr)=1/3 minimum total cost | Improved |
Closest papers by meaning
- The Effective Horizon Explains Deep RL Performance in Stochastic Environments (ICLR 2024 · similarity 0.88)
- UFO-RL: Uncertainty-Focused Optimization for Efficient Reinforcement Learning Data Selection (NeurIPS 2025 · similarity 0.87)
- Oracle-Efficient Reinforcement Learning for Max Value Ensembles (NeurIPS 2024 · similarity 0.87)
- A Differential and Pointwise Control Approach to Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Model-based RL as a Minimalist Approach to Horizon-Free and Second-Order Bounds (ICLR 2025 · similarity 0.87)
- NeoRL: Efficient Exploration for Nonepisodic RL (NeurIPS 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.
