A Primal-Dual Algorithm for Hybrid Federated Learning
Abstract
Very few methods for hybrid federated learning, where clients only hold subsets of both features and samples, exist. Yet, this scenario is very important in practical settings. We provide a fast, robust algorithm for hybrid federated learning that hinges on Fenchel Duality. We prove the convergence of the algorithm to the same solution as if the model was trained centrally in a variety of practical regimes. Furthermore, we provide experimental results that demonstrate the performance improvements of the algorithm over a commonly used method in federated learning, FedAvg, and an existing hybrid FL algorithm, HyFEM. We also provide privacy considerations and necessary steps to protect client data.
Cite
Text
Overman et al. "A Primal-Dual Algorithm for Hybrid Federated Learning." AAAI Conference on Artificial Intelligence, 2024. doi:10.1609/AAAI.V38I13.29363Markdown
[Overman et al. "A Primal-Dual Algorithm for Hybrid Federated Learning." AAAI Conference on Artificial Intelligence, 2024.](https://mlanthology.org/aaai/2024/overman2024aaai-primal/) doi:10.1609/AAAI.V38I13.29363BibTeX
@inproceedings{overman2024aaai-primal,
title = {{A Primal-Dual Algorithm for Hybrid Federated Learning}},
author = {Overman, Tom and Blum, Garrett and Klabjan, Diego},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2024},
pages = {14482-14489},
doi = {10.1609/AAAI.V38I13.29363},
url = {https://mlanthology.org/aaai/2024/overman2024aaai-primal/}
}