Research map / Reinforcement learning / Policy gradient and actor-critic
A Simple Decentralized Cross-Entropy Method
Zichen Zhang, Jun Jin, Martin Jagersand, Jun Luo, Dale Schuurmans
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 centralized CEM with multiple independent CEM instances and select the best instance | CEM planning · Continuous-control environments | evaluation return | Improved |
| Replace PETS's centralized CEM with independent CEM instances | PETS · Continuous-control environments | evaluation return | Improved |
| Use decentralized instances for action-space planning | PETS and DecentPETS · Hopper | evaluation return | No clear effect |
| Use a larger ensemble of policy networks in parameter-space planning | DecentCEM-P · Ant | evaluation return | Hurt |
| Increase ensemble size beyond the best trade-off point (ablation) | DecentCEM-A/P and DecentPETS · Reacher environments | evaluation return | Mixed |
| Increase the POPLIN policy network size to match the ensemble's total weights (ablation) | POPLIN-A/P · Reacher and PETS-Pusher | evaluation return | Mixed |
| Remove the CEM policy-improvement step (ablation) | POPLIN-A/P and DecentCEM-A/P · Continuous-control environments | evaluation return | Hurt |
| Use DecentCEM-A policy networks directly for control without CEM (ablation) | DecentCEM-A · Continuous-control environments | evaluation return | Mixed |
| Use DecentCEM-P policy networks directly for control without CEM (ablation) | DecentCEM-P · Continuous-control environments | evaluation return | Hurt |
| Increase the number of independent CEM instances at a fixed total sample budget (ablation) | DecentCEM · 1D multimodal optimization | time to run 10 trials | Hurt |
Closest papers by meaning
- DIME: Diffusion-Based Maximum Entropy Reinforcement Learning (ICML 2025 · similarity 0.86)
- Behavioral Entropy-Guided Dataset Generation for Offline Reinforcement Learning (ICLR 2025 · similarity 0.85)
- Maximum Entropy Inverse Reinforcement Learning of Diffusion Models with Energy-Based Models (NeurIPS 2024 · similarity 0.85)
- Entropy-regularized Diffusion Policy with Q-Ensembles for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.85)
- Towards Principled Unsupervised Multi-Agent Reinforcement Learning (NeurIPS 2025 · similarity 0.85)
- Natural Policy Gradient as Doubly Smoothed Policy Iteration: A Bellman-Operator Framework (NeurIPS 2026 · similarity 0.85)
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.
