Research map / Multi-agent systems and games / Cooperative multi-agent RL
QTRAN: Learning to Factorize with Transformation for Cooperative Multi-Agent 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 |
|---|---|---|---|
| Transform the joint action-value function into an additive factorization while preserving its optimal actions | QTRAN · Single-state matrix game | payoff: -3.02 / 0.03 -> 8.00 | Helped |
| Use additive factorization | VDN · Single-state matrix game | reconstructed joint action-value: 8.00 -> -3.02 | Hurt |
| Use monotonic factorization | QMIX · Single-state matrix game | reconstructed joint action-value: 8.00 -> 0.03 | Hurt |
| Replace QTRAN-base's non-optimal-action loss with the counterfactual minimum loss in QTRAN-alt (ablation) | QTRAN · Multi-domain Gaussian Squeeze | average reward | Helped |
| Replace QTRAN-base's non-optimal-action loss with the counterfactual minimum loss in QTRAN-alt (ablation) | QTRAN · Modified predator-prey, N=4 | average reward | Helped |
| Replace QTRAN-base's non-optimal-action loss with the counterfactual minimum loss in QTRAN-alt (ablation) | QTRAN · Modified predator-prey, N=4 | convergence | Helped |
| Increase the non-optimal-action penalty P | VDN and QMIX relative to QTRAN · Modified predator-prey | average reward | Mixed |
| Use full exploration without epsilon decay (ablation) | VDN and QMIX · Gaussian Squeeze | average reward | Hurt |
| Use QTRAN instead of VDN or QMIX | Value-factorization methods · 310 random 3×3 matrix games | optimal-policy learning | Helped |
Closest papers by meaning
- QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (ICML 2018 · similarity 0.91)
- Towards Understanding Cooperative Multi-Agent Q-Learning with Value Factorization (NeurIPS 2021 · similarity 0.91)
- Multiagent Q-learning with Sub-Team Coordination (NeurIPS 2022 · similarity 0.91)
- Weighted QMIX: Expanding Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (NeurIPS 2020 · similarity 0.91)
- More Centralized Training, Still Decentralized Execution: Multi-Agent Conditional Policy Factorization (ICLR 2023 · similarity 0.90)
- ConcaveQ: Non-monotonic Value Function Factorization via Concave Representations in Deep Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.89)
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.