Settling the Variance of Multi-Agent Policy Gradients

Authors: Jakub Grudzien Kuba, Muning Wen, Linghui Meng, shangding gu, Haifeng Zhang, David Mguni, Jun Wang, Yaodong Yang

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On benchmarks of Multi-Agent Mu Jo Co and Star Craft challenges, our OB technique effectively stabilises training and improves the performance of multi-agent PPO and COMA algorithms by a significant margin. Code is released at https://github.com/morning9393/ Optimal-Baseline-for-Multi-agent-Policy-Gradients.
Researcher Affiliation Collaboration 1Imperial College London, 2Huawei R&D UK, 3Shanghai Jiao Tong University, 4Institute of Automation, Chinese Academy of Science, 5University College London, 6Institute for AI, Peking University.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is released at https://github.com/morning9393/ Optimal-Baseline-for-Multi-agent-Policy-Gradients.
Open Datasets Yes Star Craft Multi-Agent Challenge (SMAC) [25]. In SMAC, each individual unit is controlled by a learning agent, which has finitely many possible actions to take. The units cooperate to defeat enemy bots across scenarios of different levels of difficulty. (...) Multi-Agent Mu Jo Co [5].
Dataset Splits No The paper does not explicitly provide specific dataset split information (exact percentages, sample counts, or detailed splitting methodology) for training, validation, and testing.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions 'PyTorch' in Appendix D but does not specify version numbers for PyTorch or any other software dependencies needed to replicate the experiment.
Experiment Setup Yes For each of the baseline on each task, we report the results of five random seeds. We refer to Appendix F for the detailed hyper-parameter settings for baselines.