EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Federated Learning with Heterogeneous Clients Using Adaptive Gradient Alignment and Partial Model Sharing

Daniel Torres, Matthew Moore

International Journal of Computational and Biological Sciences · 31.12.2025 · 2 (02) · 231-240

DOI: 10.66238/ijcbs28 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Computational and Biological Sciences, online erschienen am 31.12.2025. Verfasst von Daniel Torres und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Federated Learning (FL) has emerged as a paramount paradigm for preserving user privacy while training distributed machine learning models. However, practical deployments of FL face two critical challenges: statistical heterogeneity, often referred to as Non-IID (Independent and Identically Distributed) data, and system heterogeneity, where client devices exhibit vast disparities in computational power, battery life, and network bandwidth. Standard approaches such as Federated Averaging (FedAvg) often fail to converge or exhibit sub-optimal performance under these conditions due to weight divergence and the straggler effect. In this paper, we propose a novel framework that integrates Adaptive Gradient Alignment (AGA) with Partial Model Sharing (PMS). The AGA mechanism introduces a control variate to correct the local gradient updates, thereby mitigating the drift caused by skewed label distributions. Simultaneously, the PMS strategy dynamically adapts the model architecture transmitted to and from clients based on their real-time resource constraints, allowing participation from weak devices without stalling the global aggregation process. Extensive experiments on standard benchmark datasets demonstrate that our proposed framework achieves superior convergence rates and higher accuracy compared to state-of-the-art baselines while significantly reducing communication overhead.

Zitieren

APA

Torres, D., & Moore, M. (2025). Federated Learning with Heterogeneous Clients Using Adaptive Gradient Alignment and Partial Model Sharing. International Journal of Computational and Biological Sciences, 2(02), 231-240. https://doi.org/10.66238/ijcbs28

Vancouver

Torres D, Moore M. Federated Learning with Heterogeneous Clients Using Adaptive Gradient Alignment and Partial Model Sharing. International Journal of Computational and Biological Sciences. 2025;2(02):231-240. doi:10.66238/ijcbs28

BibTeX
@article{Torres2025,
  author = {Torres, Daniel and Moore, Matthew},
  title = {Federated Learning with Heterogeneous Clients Using Adaptive Gradient Alignment and Partial Model Sharing},
  journal = {International Journal of Computational and Biological Sciences},
  year = {2025},
  volume = {2},
  number = {02},
  pages = {231--240},
  doi = {10.66238/ijcbs28}
}
Kostenlos starten →