Research map / Reinforcement learning / Policy gradient and actor-critic
A Family of Robust Stochastic Operators for Reinforcement Learning
Yingdong Lu, Mark Squillante, Chai Wah Wu
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 the Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Acrobot | mean action gap: 0.1436 -> 0.9004 | Improved |
| Replace the Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Mountain Car test | average score: 129.93 -> 122.70 | Improved |
| Replace the Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Mountain Car test | score standard deviation: 32.68 -> 7.25 | Improved |
| Replace the Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Cart Pole, last 100 episodes | average score: 183.67 -> 190.92 | Improved |
| Replace the Consistent Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Cart Pole, last 100 episodes | average score: 184.07 -> 190.92 | Improved |
| Replace the Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Cart Pole | score standard deviation: 28.87 -> 19.44 | Improved |
| Replace the Bellman operator with RSO using βₖ ~ U[0,2) | Q-learning · Lunar Lander test | average score: -241.94 -> -167.51 | Improved |
| Use βₖ ~ U[0,1) instead of βₖ ~ U[0,2) (ablation) | RSO · OpenAI Gym problems | performance | Hurt |
| Fix βₖ = 1 instead of sampling βₖ ~ U[0,2) (ablation) | RSO · OpenAI Gym problems | performance | Hurt |
| Use mixtures of β distributions across iterations rather than βₖ ~ U[0,2) throughout (ablation) | RSO · OpenAI Gym problems | performance | Hurt |
| Increase variability in the βₖ sequence while preserving its mean | Robust stochastic operators · Theoretical action-gap result | action gap | Improved |
Closest papers by meaning
- Finite-Sample Analysis of Policy Evaluation for Robust Average Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Efficient Value Iteration for s-rectangular Robust Markov Decision Processes (ICML 2024 · similarity 0.86)
- State Entropy Regularization for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Adversarial Diffusion for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Time-Constrained Robust MDPs (NeurIPS 2024 · similarity 0.86)
- Minimax Optimal and Computationally Efficient Algorithms for Distributionally Robust Offline Reinforcement Learning (NeurIPS 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.
