UWMMSE

November 2, 2020 ยท View on GitHub

Tensorflow implementation of Unfolding WMMSE using Graph Neural Networks for Efficient Power Allocation (https://arxiv.org/abs/2009.10812)

Overview

This library contains a Tensorflow implementation of Unfolding WMMSE using Graph Neural Networks for Efficient Power Allocation as presented in [1](https://arxiv.org/abs/2009.10812).

Dependencies

Structure

  • datagen: Code to generate dataset. Generates A.pkl ( Geometric graph ), H.pkl ( Dictionary containing train_H and test_H ) and coordinates.pkl ( node position coordinates ). Run as python3 datagen.py [dataset ID]. User chosen [dataset ID] will be used as the foldername to store dataset. Eg., to generate dataset with ID set3, run python3 datagen.py set3.
  • data: should contain your dataset in folder [dataset ID].
  • main: Main code for running the experiments in the paper. Run as python3 main.py [dataset ID] [exp ID] [mode]. Eg., to train UWMMSE on dataset with ID set3, run python3 main.py set3 uwmmse train.
  • model: Defines the UWMMSE model.
  • models: Stores trained models in a folder with same name as [dataset ID].

Usage

Please cite [1] in your work when using this library in your experiments.

Feedback

For questions and comments, feel free to contact Arindam Chowdhury.

Citation

[1] Chowdhury A, Verma G, Rao C, Swami A, Segarra S. Unfolding WMMSE using Graph Neural Networks 
for Efficient Power Allocation. arXiv preprint arXiv:2009.10812. 2020 Sep 22.

BibTeX format:

@article{chowdhury2020unfolding,
  title={Unfolding WMMSE using Graph Neural Networks for Efficient Power Allocation},
  author={Chowdhury, Arindam and Verma, Gunjan and Rao, Chirag and Swami, Ananthram and Segarra, Santiago},
  journal={arXiv preprint arXiv:2009.10812},
  year={2020}
}