Research map / Multi-agent systems and games / Cooperative multi-agent RL
MAP Propagation Algorithm: Faster Learning with a Team of Reinforcement Learning Agents
Stephen Chung
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 |
|---|---|---|---|
| Use MAP propagation instead of REINFORCE | Team of agents · Acrobot-v1 | average return: -148.42 -> -100.29 | Improved |
| Use MAP propagation instead of REINFORCE | Team of agents · CartPole-v1 | average return: 47.29 -> 459.70 | Improved |
| Use MAP propagation instead of REINFORCE | Team of agents · LunarLander-v2 | average return: -62.05 -> 127.88 | Improved |
| Use MAP propagation instead of REINFORCE | Team of agents · MountainCarContinuous-v0 | average return: -35.52 -> 39.45 | Improved |
| Use MAP propagation instead of backprop | Actor-critic network · Acrobot-v1 | average return: -106.29 -> -100.29 | Improved |
| Use MAP propagation instead of backprop | Actor-critic network · CartPole-v1 | average return: 458.96 -> 459.70 | Improved |
| Use MAP propagation instead of backprop | Actor-critic network · LunarLander-v2 | average return: 104.92 -> 127.88 | Improved |
| Use MAP propagation instead of backprop | Actor-critic network · MountainCarContinuous-v0 | average return: 4.30 -> 39.45 | Improved |
| Add target-output information to the MAP learning rule (ablation) | MAP propagation on scalar regression · Scalar regression | learning speed | Improved |
| Use negative L2 loss as reward rather than the target-informed Appendix C rule (ablation) | MAP propagation on scalar regression · Scalar regression | learning speed | Hurt |
| Use target-informed MAP propagation rather than backprop | Scalar regression · Scalar regression | asymptotic performance | Hurt |
Closest papers by meaning
- Attention-Gated Brain Propagation: How the brain can implement reward-based error backpropagation (NeurIPS 2020 · similarity 0.88)
- Efficient Reinforcement Learning by Discovering Neural Pathways (NeurIPS 2024 · similarity 0.87)
- MAPF-GPT: Imitation Learning for Multi-Agent Pathfinding at Scale (AAAI 2025 · similarity 0.87)
- Stay Hungry, Keep Learning: Sustainable Plasticity for Deep Reinforcement Learning (ICML 2025 · similarity 0.85)
- Learn to Follow: Decentralized Lifelong Multi-Agent Pathfinding via Planning and Learning (AAAI 2024 · similarity 0.85)
- Bridging Training and Execution via Dynamic Directed Graph-Based Communication in Cooperative Multi-Agent Systems (AAAI 2025 · 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.
