Stabilizing Federated Learning under Extreme Heterogeneity with HeteRo-Select
Published in arXiv Preprint, 2025
This paper presents HeteRo-Select, a novel approach to stabilize federated learning under extreme data heterogeneity scenarios. The method addresses challenges in federated learning where client data distributions vary significantly, proposing a robust selection mechanism for improved model convergence and performance.
| Download paper here | PDF Version |
Research Area: Federated Learning, Machine Learning, Distributed Systems
Recommended citation: Masud, M. A., Jahin, M. A., & Hasan, M. (2025). Stabilizing Federated Learning under Extreme Heterogeneity with HeteRo-Select. arXiv preprint arXiv:2508.06692.
Download Paper