FedEFC: Federated Learning Using Enhanced Forward Correction Against Noisy Labels
#federated learning #noisy labels #forward correction #heterogeneous data #communication constraints #privacy preservation #distributed training
📌 Key Takeaways
- FedEFC extends forward correction to federated learning to mitigate noisy labels.
- The method adapts to client‑specific data heterogeneity and limited communication budgets.
- Experimental evaluations demonstrate that FedEFC consistently outperforms existing FL approaches on noisy datasets.
- The approach preserves client privacy by operating on model updates only, without revealing raw data.
📖 Full Retelling
In April 2025, a group of researchers announced FedEFC, a new federated learning method that tackles the long‑standing problem of noisy labels in distributed, privacy‑preserving model training. FedEFC is designed to address performance degradation caused by both heterogeneous data distributions across clients and communication constraints, and it does so by extending the forward correction technique—traditionally used in centralized settings—to the federated learning environment.
🏷️ Themes
Federated Learning, Robustness to Noisy Labels, Forward Correction Techniques, Privacy‑Preserving Distributed Training, Communication Efficiency in FL
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
arXiv:2504.05615v3 Announce Type: replace-cross
Abstract: Federated Learning (FL) is a powerful framework for privacy-preserving distributed learning. It enables multiple clients to collaboratively train a global model without sharing raw data. However, handling noisy labels in FL remains a major challenge due to heterogeneous data distributions and communication constraints, which can severely degrade model performance. To address this issue, we propose FedEFC, a novel method designed to tackl
Read full article at source