Research map / Reinforcement learning / MDPs and planning
Semi-Infinitely Constrained Markov Decision Processes and Provably Efficient Reinforcement Learning
Liangyu Zhang, Peng Yang, Wenhao Yang, Zhihua 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 |
|---|---|---|---|
| solve with SI-CRL (exchange method over continuum of constraints) instead of naive discretization into a CMDP | tabular CMDP solver · Discharge of Sewage | solution quality | Improved |
| use SI-CRL with dual simplex re-optimization instead of a single LP over a discretized grid | tabular CMDP solver · Discharge of Sewage | time to reach given accuracy | Improved |
| use SI-CRL instead of discretization baseline | tabular CMDP solver · Discharge of Sewage | number of active constraints | Improved |
| use SI-CPO instead of discretization baselines solved by CRPO | policy optimization · Discharge of Sewage | convergence speed | Improved |
| use SI-CPO instead of discretization baseline with N=1000 | policy optimization · Discharge of Sewage | time per iteration (s): 0.45 ± 0.04 -> 0.19 ± 0.02 | Improved |
| increase discretization grid size N_baseline | CRPO baseline · Ship Route Planning | convergence | No clear effect |
| use SI-CPPO instead of discretization baseline | policy optimization · Ship Route Planning | route feasibility | Improved |
| use SI-CPPO instead of discretization baseline N=250 | policy optimization · Ship Route Planning | time per iteration (s): 8.91 ± 1.50 -> 11.39 ± 1.43 | Hurt |
| use SI-CPPO instead of discretization baseline | policy optimization · Ship Route Planning | convergence | Improved |
Closest papers by meaning
- Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation (NeurIPS 2025 · similarity 0.90)
- Learning General Parameterized Policies for Infinite Horizon Average Reward Constrained MDPs via Primal-Dual Policy Gradient Algorithm (NeurIPS 2024 · similarity 0.90)
- Achieving $\tilde{O}(1/\epsilon)$ Sample Complexity for Constrained Markov Decision Process (NeurIPS 2024 · similarity 0.90)
- Polynomial-Time Approximability of Constrained Reinforcement Learning (ICML 2025 · similarity 0.89)
- Confident Natural Policy Gradient for Local Planning in $q_\pi$-realizable Constrained MDPs (NeurIPS 2024 · similarity 0.89)
- Global Convergence for Average Reward Constrained MDPs with Primal-Dual Actor Critic Algorithm (NeurIPS 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.
