Reproduce “Communication-Efficient Hierarchical Federated Learning for IoT Heterogeneous Systems with Imbalanced Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Alaa Awad Abdellatif, Naram Mhaisen, Amr Mohamed, Aiman Erbad, Mohsen Guizani, Zaher Dawy, Wassim Nasreddine
This research introduces an optimized hierarchical federated learning system for heterogeneous IoT environments with imbalanced data, evaluated using real-world healthcare datasets. The proposed method achieves a 75-85% reduction in communication rounds and up to a 6% increase in classification accuracy compared to distance-based federated learning, while also reducing communication traffic by up to 73%.
Key insights
- Intelligently assigning end-users (EUs) to edge nodes based on their data distribution, rather than mere proximity, significantly improves model accuracy and accelerates convergence in non-IID data settings.
- The Dual Connectivity Allocation (DCA) strategy, allowing EUs to connect to multiple edge nodes, further enhances data balancing and can enable hierarchical FL to outperform centralized learning benchmarks.
- Joint optimization of EU assignment and resource allocation effectively addresses practical communication and computational constraints, making federated learning more viable for resource-constrained IoT healthcare systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.