Learning to Share in Networked Multi-Agent Reinforcement Learning

Authors: Yuxuan Yi, Ge Li, Yaowei Wang, Zongqing Lu

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically demonstrate that LTo S outperforms existing methods in both social dilemma and networked MARL scenarios across scales.
Researcher Affiliation Academia 1Peking University 2Peng Cheng Lab
Pseudocode Yes For completeness, Algorithm 1 (see Appendix B) gives the training procedure of LTo S based on DDPG and DGN.
Open Source Code Yes The code of LTo S is available at https://github.com/PKU-RL/Roadnet SZ.
Open Datasets Yes We adopt the same problem setting as Wei et al. (2019). In a road network, each agent serves as traffic signal control at an intersection. The experiment was conducted on a traffic simulator, City Flow (Zhang et al., 2019). We use a 6 6 grid network with 36 intersections. The traffic flows were generated to simulate dynamic traffic flows including both peak and off-peak period, and the statistics are summarized in Table 2.
Dataset Splits No The paper mentions '5 training runs with different random seeds' but does not specify any training/validation/test dataset splits for reproducibility.
Hardware Specification No The paper states 'All our experiments do not claim for much computation.' but does not provide any specific details about the hardware (e.g., GPU models, CPU types) used for the experiments.
Software Dependencies No The paper mentions using DDPG and DGN, but does not provide specific version numbers for these libraries or any other software dependencies like Python, PyTorch, or TensorFlow versions.
Experiment Setup No The paper states 'More details of hyperparameters are available in Appendix D.' but does not provide these details within the main text of the paper.