Dynamic SVD-Enhanced Approach for Federated Learning

Jianbo Zhang, Lena Mashayekhy

Keywords: Federated Learning

Abstract: Federated Learning (FL) has emerged as a promising paradigm for collaborative machine learning while preserving data privacy. However, existing FL approaches face challenges in balancing model generalization among heterogeneous clients and resistance to malicious attacks. This paper introduces Dynamic SVD-driven Federated Learning (DSVD-FL), a novel approach that addresses these challenges simultaneously. DSVD-FL dynamically adjusts the contribution of each client using Singular Value Decomposition (SVD), introducing an adaptive weighting mechanism based on singular value contributions and vector alignments. Theoretical analysis demonstrates the convergence properties and computational efficiency of our approach. Experimental results on both IID and non-IID datasets show that DSVD-FL outperforms state-of-the-art FL approaches in terms of model accuracy, robustness against various attack scenarios, while maintaining competitive computational efficiency. We perform an ablation study to explore the key components of SVD that impact the federated learning performance.

Primary Area: alignment, fairness, safety, privacy, and societal considerations

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 3264