<ul><li>Federated Learning (FL) is a method for collaborative learning among edge devices by sharing models instead of raw data.</li><li>A novel method called Sharpness-Aware Minimization-based Robust Federated Learning (SMRFL) has been proposed to improve model robustness against perturbations.</li><li>SMRFL minimizes the maximum loss within a neighborhood of model parameters to reduce sensitivity to perturbations and enhance robustness.</li><li>Experimental results show that SMRFL enhances robustness against perturbations on real-world datasets compared to three baseline methods.</li></ul>

Robust Federated Learning against Model Perturbation in Edge Networks

Discover more