Research map / Reinforcement learning / Policy gradient and actor-critic
Quantum Robust Inner Minimization for Reinforcement Learning with Quadratic Speed-Up in Query Complexity
Hyun Kyu Lee, Joongheon Kim, Sung Whan Yoon
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 |
|---|---|---|---|
| replace exhaustive scan with quantum minimum finding (QRIM) in inner minimization | Classical RRL · CartPole Swingup extrapolation | mean discounted return: 39.86 -> 58.06 | Improved |
| replace exhaustive scan with QRIM | Classical RRL · FrozenLake-8x8 extrapolation | mean discounted return: −0.45 -> −0.37 | Improved |
| replace exhaustive scan with QRIM | Classical RRL · FrozenLake-8x8 train | mean discounted return: 0.84 -> 0.98 | Improved |
| replace exhaustive scan with QRIM | Classical RRL · CartPole Swingup train | mean discounted return: 61.03 -> 76.41 | Improved |
| replace exhaustive scan with QRIM | Classical RRL · FrozenLake-8x8 | cumulative oracle calls: 11.45M -> 4.11M | Improved |
| replace exhaustive scan with QRIM | Classical RRL · CartPole Swingup | cumulative oracle calls: 27.51M -> 5.65M | Improved |
| add robust training over uncertainty set | Classic RL · FrozenLake-8x8 extrapolation | mean discounted return: −0.79 -> −0.37 | Improved |
| replace BAI-based inner search with QRIM | BAI-RRL · FrozenLake-8x8 | cumulative oracle calls: 9.67M -> 4.11M | Improved |
| run QRIM on real quantum hardware instead of noise-free simulation | QRIM · FrozenLake-8x8 extrapolation | mean discounted return: −0.37 -> −0.52 | Hurt |
| increase QAE tolerance from 0.01 to 0.20 (ablation) | QRIM · CartPole Swingup train | mean evaluation return: 76.41 -> 56.14 | Hurt |
| increase QAE tolerance from 0.01 to 0.20 (ablation) | QRIM · CartPole Swingup extrapolation | mean evaluation return: 58.06 -> 27.04 | Hurt |
| increase QAE tolerance from 0.01 to 0.10 (ablation) | QRIM · CartPole Swingup train | mean evaluation return: 76.41 -> 72.96 | No clear effect |
Closest papers by meaning
- Provably Efficient Exploration in Quantum Reinforcement Learning with Logarithmic Worst-Case Regret (ICML 2024 · similarity 0.88)
- Accelerating Quantum Reinforcement Learning with a Quantum Natural Policy Gradient Based Approach (ICML 2025 · similarity 0.87)
- Benchmarking Quantum Reinforcement Learning (ICML 2025 · similarity 0.86)
- Reinforcement Learning for Quantum Control under Physical Constraints (ICML 2025 · similarity 0.86)
- Towards Optimal Adversarial Robust Q-learning with Bellman Infinity-error (ICML 2024 · similarity 0.86)
- Zero-Sum Positional Differential Games as a Framework for Robust Reinforcement Learning: Deep Q-Learning Approach (ICML 2024 · similarity 0.86)
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.
