Research map / Multi-agent systems and games / Cooperative multi-agent RL
Non-Linear Coordination Graphs
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 |
|---|---|---|---|
| Replace linear (conventional) value decomposition with a non-linear LeakyReLU mixing network over coordination graph | DCG · Two-step matrix game, State A | learned Q of Action B: 5.85 -> 7.95 | Helped |
| Replace linear value decomposition with non-linear mixing | DCG · Two-step matrix game, State 2B | learned Q with all agents on Action B: 5.89 -> 8.02 (true 8) | Helped |
| Use non-linear coordination graph on complex coordination tasks | DCG/QMIX/DICG · MACO Pursuit | return | Helped |
| Fully decomposed value (QMIX) instead of coordination graph | NL-CG · MACO Pursuit | return | Hurt |
| Increase mixing network hidden width from 2, 3, 4 to 10 (ablation) | NL-CG · MACO Pursuit | return | Helped |
| Iterative optimization (width 4, n_max=4) instead of enumerating all configurations (width 3) (ablation) | NL-CG · MACO Pursuit | return | No clear effect |
| Iterative method with hidden width 10 instead of enumerative method, both checking 16 pieces (ablation) | NL-CG · MACO | return: 16 linear pieces checked | Helped |
| Iterative instead of enumerative action selection (ablation) | NL-CG · MACO | running time: saves 50% to 65% | Helped |
| Iterative method after 4 slope configurations vs enumeration (ablation) | NL-CG · MACO | Q_tot of selected actions | No clear effect |
| Estimate Q_tot with NL-CG versus ground truth | NL-CG · MACO | estimation error: less than 20% | Helped |
Closest papers by meaning
- Self-Organized Polynomial-Time Coordination Graphs (ICML 2022 · similarity 0.92)
- Deep Coordination Graphs (ICML 2020 · similarity 0.90)
- Greedy based Value Representation for Optimal Coordination in Multi-agent Reinforcement Learning (ICML 2022 · similarity 0.89)
- Graph Diffusion for Robust Multi-Agent Coordination (ICML 2025 · similarity 0.88)
- Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.87)
- ConcaveQ: Non-monotonic Value Function Factorization via Concave Representations in Deep Multi-Agent Reinforcement Learning (AAAI 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.