FedRG: Unleashing the Representation Geometry for Federated Learning with Noisy Clients

1. 1Rethink the paradigm of recognizing noisy samples from scalar loss values to representation geometry
2. 2Leverage representation geometry to improve the reliability of noisy samples recognition in heterogeneous scenarios
3. 3Apply FedRG to federated learning with noisy clients to enhance model performance

Machine learning researchers and engineers working on federated learning benefit from this approach as it enhances model performance in distributed scenarios with noisy annotations

💡 Representation geometry can be used to improve the reliability of noisy samples recognition in federated learning

🚀 FedRG boosts federated learning with noisy clients by unlocking representation geometry!