Efficient Episodic Memory Utilization of Cooperative Multi-Agent Reinforcement Learning

Authors: Hyungho Na, Yunkyeong Seo, Il-chul Moon

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed method is evaluated in Star Craft II and Google Research Football, and empirical results indicate further performance improvement over state-of-the-art methods.
Researcher Affiliation Collaboration Hyungho Na1, Yunkyeong Seo1 & Il-Chul Moon1,2 1Korea Advanced Institute of Science and Technology (KAIST), 2summary.ai
Pseudocode Yes Algorithm 1 Training Algorithm for State Embedding Algorithm 3 EMU: Efficient episodic Memory Utilization for MARL
Open Source Code Yes Our code is available at: https://github.com/Hyungho Na/EMU.
Open Datasets Yes We conduct experiments on complex multi-agent tasks such as SMAC (Samvelyan et al., 2019) and GRF (Kurach et al., 2020).
Dataset Splits No The paper mentions training and testing but does not explicitly describe a separate 'validation' dataset split with specific percentages or counts.
Hardware Specification Yes Experiments for SMAC (Samvelyan et al., 2019) are mainly carried out on NVIDIA Ge Force RTX 3090 GPU
Software Dependencies No We utilize Py MARL (Samvelyan et al., 2019) to execute all of the baseline algorithms with their open-source codes... All SMAC experiments were conducted on Star Craft II version 4.10.0 in a Linux environment. The paper mentions these software components but does not provide version numbers for all key dependencies like Python, PyTorch, etc.
Experiment Setup Yes Table 1: Task-dependent hyperparameter of EMU. Table 8: EMU Hyperparameters for fϕ and fψ training.