Federated Edge Learning: Design Issues and Challenges

Federated Learning (FL) is a distributed machine learning technique, where each device contributes to the model by independently computing gradient based on its local training data. It has recently become hot research topic, as it promises several benefits related data privacy and scalability. However, implementing FL at network edge challenging due system heterogeneity resource constraints. In this article, we examine existing challenges trade-offs in Edge (FEEL). The design of FEEL algorithms for resources-efficient raises challenges. These are essentially multidisciplinary nature problem. As key component learning, article advocates new set considerations characteristics wireless scheduling FEEL. Hence, propose general framework data-aware guideline future directions. We also discuss main axes requirements evaluation some exploitable techniques metrics.