Research map / Reinforcement learning / Policy gradient and actor-critic
Is Model Ensemble Necessary? Model-based RL via a Single Model with Lipschitz Regularized Value Function
Ruijie Zheng, Xiyao Wang, Huazhe Xu, Furong Huang
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 |
|---|---|---|---|
| Add robust regularization to the value network while using a single deterministic dynamics model | MBPO · Walker, HalfCheetah, Ant, Humanoid, Hopper | return | Improved |
| Use robust regularization instead of a probabilistic model ensemble | MBPO · Walker | wall-clock time: 60.0 -> 45.4 h | Improved |
| Use robust regularization instead of a probabilistic model ensemble | MBPO · Ant | wall-clock time: 62.3 -> 48.3 h | Improved |
| Add random uniform noise instead of adversarially selecting perturbations (ablation) | Robust Regularization on MBPO · Walker | return | Hurt |
| Use PGD-20 instead of FGSM to compute perturbations (ablation) | Robust Regularization on MBPO · Walker | return | No clear effect |
| Vary the robust-regularization weight λ (ablation) | Robust Regularization on MBPO · Walker | return | Mixed |
| Use a tighter global spectral-normalization constraint (ablation) | Spectral Normalization on MBPO · Walker | return | Mixed |
| Combine spectral normalization with a single probabilistic model rather than a single deterministic model (ablation) | MBPO · Ant | return | Improved |
| Combine robust regularization with a single probabilistic model rather than a single deterministic model (ablation) | MBPO · Ant | return | Improved |
| Add Lipschitz regularization to model-free SAC (ablation) | SAC · MuJoCo tasks | return | Improved |
Closest papers by meaning
- Sample-Efficient Reinforcement Learning with Stochastic Ensemble Value Expansion (NeurIPS 2018 · similarity 0.89)
- Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models (NeurIPS 2018 · similarity 0.88)
- The Edge-of-Reach Problem in Offline Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Enhancing Value Function Estimation through First-Order State-Action Dynamics in Offline Reinforcement Learning (ICML 2024 · similarity 0.86)
- Learning to Trust Bellman Updates: Selective State-Adaptive Regularization for Offline RL (ICML 2025 · similarity 0.86)
- Provable Risk-Sensitive Distributional Reinforcement Learning with General Function Approximation (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.
