Research map / Reinforcement learning / Policy gradient and actor-critic
Adaptive Trust Region Policy Optimization: Global Convergence and Faster Rates for Regularized MDPs
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 adaptive-scaled trust-region updates in Uniform TRPO | Tabular discounted unregularized MDPs · Tabular discounted MDP | value error convergence rate: \tilde O(1/\sqrt{N}) | Helped |
| Add strongly convex instantaneous-cost regularization | Uniform TRPO · Tabular discounted regularized MDP | value error convergence rate: \tilde O(1/\sqrt{N}) -> \tilde O(1/N) | Helped |
| Use sample-based adaptive-scaled trust-region updates | Sample-Based TRPO · Sample-based unregularized MDP | best-policy value error convergence rate: \tilde O(1/\sqrt{N}) | Helped |
| Add strongly convex instantaneous-cost regularization | Sample-Based TRPO · Sample-based regularized MDP | best-policy value error convergence rate: \tilde O(1/\sqrt{N}) -> \tilde O(1/N) | Helped |
| Use Sample-Based TRPO instead of CPI | Sample-based unregularized policy search · Sample-based MDP | sample complexity dependence on (1-γ)^{-1}: (1-\gamma)^{-5} -> (1-\gamma)^{-3} | Helped |
| Use regularized Sample-Based TRPO rather than its unregularized version | Sample-Based TRPO · Sample-based MDP | sample complexity dependence on ε: \epsilon^{-4} -> \epsilon^{-3} | Helped |
| Remove improvement-based updates and early stopping | Sample-Based TRPO · Sample-based MDP | sample complexity dependence on (1-γ)^{-1}: (1-\gamma)^{-5} -> (1-\gamma)^{-3} | Helped |
Closest papers by meaning
- Simple Policy Optimization (ICML 2025 · similarity 0.93)
- ACPO: A Policy Optimization Algorithm for Average MDPs with Constraints (ICML 2024 · similarity 0.89)
- Reflective Policy Optimization (ICML 2024 · similarity 0.89)
- Embedding Safety into RL: A New Take on Trust Region Methods (ICML 2025 · similarity 0.89)
- Finite-Sample Convergence Bounds for Trust Region Policy Optimization in Mean Field Games (ICML 2025 · similarity 0.88)
- Trust Region Reward Optimization and Proximal Inverse Reward Optimization Algorithm (NeurIPS 2025 · 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.