Federated Learning in IoT

Distributed Learning technique improved the training model performance, while its development is at the cost of sensitive data exposure during centralized training process. Federated Learning (FL) thus emerged and served as an enabling solution to bridge the gap without sacrificing privacy, however, it is inevitably vulnerable to heterogeneity problems. Rely on the benefits to this heterogeneity characteristic and decrease the risk in the collaborative training, we introduce the self-organizing concept to existing FL settings to create a more intelligent central server to improve the performance of FL.