Graph Convolutional Reinforcement Learning

Authors: Jiechuan Jiang, Chen Dun, Tiejun Huang, Zongqing Lu

ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we show that our method substantially outperforms existing methods in a variety of cooperative scenarios. For the experiments, we adopt a grid-world platform MAgent (Zheng et al., 2017).
Researcher Affiliation Collaboration 1Peking University, 2Rice University. This work was supported in part by NSF China under grant 61872009, Huawei Noah s Ark Lab, and Peng Cheng Lab.
Pseudocode No No pseudocode or algorithm blocks found.
Open Source Code Yes The code of DGN is available at https://github.com/PKU-AI-Edge/DGN/.
Open Datasets Yes For the experiments, we adopt a grid-world platform MAgent (Zheng et al., 2017). The paper describes three experimental scenarios: battle, jungle, and routing, which are built environments within the MAgent platform.
Dataset Splits No No explicit mention of a 'validation' split or dataset was found. The paper describes training and testing phases but does not explicitly specify a validation set for hyperparameter tuning or early stopping.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for experiments are provided in the paper.
Software Dependencies No No specific software dependencies with version numbers are provided in the paper. It mentions 'Tensor Flow' but without a version.
Experiment Setup Yes Table 4: Hyperparameters in Appendix A summarizes detailed hyperparameters used by DGN and the baselines, including discount (γ), batch size, buffer capacity, β, ϵ and decay, optimizer (Adam), learning rate, # neighbors, # convolutional layers, # attention heads, τ, λ, κ, # encoder MLP layers and units, Q network type, MLP activation, and initializer.