FedClean: A General Robust Label Noise Correction for Federated Learning

Xiaoqian Jiang, Jing Zhang
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:27772-27792, 2025.

Abstract

Many federated learning scenarios encounter label noises in the client-side datasets. The resulting degradation in global model performance raises the urgent need to address label noise. This paper proposes FedClean – a novel general robust label noise correction for federated learning. FedClean first uses the local centralized noisy label learning to select clean samples to train a global model. Then, it employs a two-stage correction scheme to correct the noisy labels from two distinct perspectives of local noisy label learning and the global model. FedClean also proposes a novel model aggregation method, further reducing the impact of label noises. FedClean neither assumes the existence of clean clients nor the specific noise distributions, showing the maximum versatility. Extensive experimental results show that FedClean effectively identifies and rectifies label noises even if all clients exhibit label noises, which outperforms the state-of-the-art noise-label learning methods for federated learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-jiang25m, title = {{F}ed{C}lean: A General Robust Label Noise Correction for Federated Learning}, author = {Jiang, Xiaoqian and Zhang, Jing}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {27772--27792}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/jiang25m/jiang25m.pdf}, url = {https://proceedings.mlr.press/v267/jiang25m.html}, abstract = {Many federated learning scenarios encounter label noises in the client-side datasets. The resulting degradation in global model performance raises the urgent need to address label noise. This paper proposes FedClean – a novel general robust label noise correction for federated learning. FedClean first uses the local centralized noisy label learning to select clean samples to train a global model. Then, it employs a two-stage correction scheme to correct the noisy labels from two distinct perspectives of local noisy label learning and the global model. FedClean also proposes a novel model aggregation method, further reducing the impact of label noises. FedClean neither assumes the existence of clean clients nor the specific noise distributions, showing the maximum versatility. Extensive experimental results show that FedClean effectively identifies and rectifies label noises even if all clients exhibit label noises, which outperforms the state-of-the-art noise-label learning methods for federated learning.} }
Endnote
%0 Conference Paper %T FedClean: A General Robust Label Noise Correction for Federated Learning %A Xiaoqian Jiang %A Jing Zhang %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-jiang25m %I PMLR %P 27772--27792 %U https://proceedings.mlr.press/v267/jiang25m.html %V 267 %X Many federated learning scenarios encounter label noises in the client-side datasets. The resulting degradation in global model performance raises the urgent need to address label noise. This paper proposes FedClean – a novel general robust label noise correction for federated learning. FedClean first uses the local centralized noisy label learning to select clean samples to train a global model. Then, it employs a two-stage correction scheme to correct the noisy labels from two distinct perspectives of local noisy label learning and the global model. FedClean also proposes a novel model aggregation method, further reducing the impact of label noises. FedClean neither assumes the existence of clean clients nor the specific noise distributions, showing the maximum versatility. Extensive experimental results show that FedClean effectively identifies and rectifies label noises even if all clients exhibit label noises, which outperforms the state-of-the-art noise-label learning methods for federated learning.
APA
Jiang, X. & Zhang, J.. (2025). FedClean: A General Robust Label Noise Correction for Federated Learning. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:27772-27792 Available from https://proceedings.mlr.press/v267/jiang25m.html.

Related Material