C/D-ADMM

This repository is the simplest implementation for the paper 'Online Parallel Multi-Task Relationship Learning via Alternating Direction Method of Multipliers', accepted by Neurocomputing Journal.

Guidance

data/synthetic.npz: the synthetic dataset used in the paper.

src/utils.py: some commonly used functions.

src/network.py: generate different topologies with networkx package.

src/gendata.py: generate the synthetic data from scratch.

src/single_task.py: the ADMM-Single baseline compared in the paper.

src/multi_task.py: the C-ADMM algorithm proposed in the paper.

src/admm.py: the D-ADMM algorithm proposed in the paper.

Citation

Please cite our paper if this paper/code helps your research.

@article{ADMM2024,
title = {Online parallel multi-task relationship learning via alternating direction method of multipliers},
journal = {Neurocomputing},
volume = {611},
pages = {128702},
year = {2025},
issn = {0925-2312},
doi = {https://doi.org/10.1016/j.neucom.2024.128702},
url = {https://www.sciencedirect.com/science/article/pii/S0925231224014735},
author = {Ruiyu Li and Peilin Zhao and Guangxia Li and Zhiqiang Xu and Xuewei Li}
}