Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Graph Convolutional Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 temporal relation regularization (ablation) | DGN · Battle | mean reward: 0.84 -> 0.91 | Helped |
| replace mean kernels with relation kernels (ablation) | DGN-M · Battle | mean reward: 0.50 -> 0.84 | Helped |
| use graph convolution rather than parameter sharing without graph convolution (ablation) | cooperative multi-agent methods · Battle | mean reward | Helped |
| increase the number of convolutional layers from one to two (ablation) | DGN · Battle | learning stability | Helped |
| increase neighbor count from 1 to 3 (ablation) | DGN · Jungle | performance | Helped |
| increase neighbor count from 3 to 4 (ablation) | DGN · Jungle | performance | Hurt |
| hold the agent graph fixed across two successive timesteps during training (ablation) | DGN · Battle | learning convergence speed | Helped |
| use a multi-head attention relation kernel instead of an LSTM kernel (ablation) | DGN · Battle | performance | Helped |
| restrict communication to graph neighbors instead of full communication (ablation) | DGN · Battle | performance | Helped |
| add temporal relation regularization (ablation) | DGN · Jungle | performance | No clear effect |
| add temporal relation regularization (ablation) | DGN · Routing | performance | Helped |
| apply a policy trained with N=20 to larger networks without retraining | DGN · Routing (N=40 and N=60, L=20) | mean reward: 0.86 vs 0.78; 0.73 vs 0.59 | Helped |
Closest papers by meaning
- Towards Efficient Collaboration via Graph Modeling in Reinforcement Learning (AAAI 2025 · similarity 0.89)
- Bridging Training and Execution via Dynamic Directed Graph-Based Communication in Cooperative Multi-Agent Systems (AAAI 2025 · similarity 0.87)
- Multi-agent cooperation through learning-aware policy gradients (ICLR 2025 · similarity 0.87)
- Enhancing Cooperative Multi-Agent Reinforcement Learning with State Modelling and Adversarial Exploration (ICML 2025 · similarity 0.87)
- Graph Diffusion for Robust Multi-Agent Coordination (ICML 2025 · similarity 0.87)
- HYGMA: Hypergraph Coordination Networks with Dynamic Grouping for Multi-Agent Reinforcement Learning (ICML 2025 · 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.