Research map / Reinforcement learning / Visual and representation RL
Improving Generalization in Reinforcement Learning with Mixture Regularization
KAIXIN WANG, Bingyi Kang, Jie Shao, Jiashi Feng
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 |
|---|---|---|---|
| Train PPO with mixreg, mixing observations and their supervision signals | PPO · Procgen 500-level generalization | zero-shot testing performance | Improved |
| Train Rainbow with mixreg, mixing observations and their supervision signals | Rainbow · Procgen 500-level generalization | zero-shot testing performance | Improved |
| Combine mixreg with other methods | PPO on Starpilot · Procgen Starpilot, 500-level generalization | testing performance | Improved |
| Use cutout-color, random crop, or random convolution instead of mixreg | PPO · Procgen 500-level generalization | zero-shot testing performance | Mixed |
| Add mixreg | PPO with different model sizes · Procgen 500-level generalization | zero-shot testing performance | Improved |
| Add mixreg | PPO on Fruitbot · Procgen Fruitbot, 500-level generalization | sample efficiency | Improved |
| Mix observations without mixing the associated supervision signals (ablation) | PPO mixreg · Procgen games | zero-shot testing performance | No clear effect |
| Add interpolation of the associated supervision signals to observation mixing (ablation) | PPO mixreg · Procgen games | policy generalizability | Improved |
| Use a larger beta-distribution parameter, such as α=1.0, to sample mixing coefficients (ablation) | Mixreg · Procgen environments | testing performance | Hurt |
Closest papers by meaning
- Quantifying Generalization in Reinforcement Learning (ICML 2019 · similarity 0.89)
- Generalization in Reinforcement Learning with Selective Noise Injection and Information Bottleneck (NeurIPS 2019 · similarity 0.88)
- On the Power of Pre-training for Generalization in RL: Provable Benefits and Hardness (ICML 2023 · similarity 0.88)
- Regularization Matters in Policy Optimization - An Empirical Study on Continuous Control (ICLR 2021 · similarity 0.88)
- How Ensembles of Distilled Policies Improve Generalisation in Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Doubly Mild Generalization for 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.
