Edge Intelligent Computing Enabled Federated Learning in 6G Wireless Systems

Federated Learning (FL) methodologies are expected to enable a huge number of applications according to intelligent distributed frameworks based on machine learning. In such a context, the sixth-generation (6G) networks technology appears as a promising opportunity to offer fast and reliable communications. This chapter illustrates a FL framework that takes into account both the hesitation of users to take part at the Fl process without receiving compensation and the impact of the communication channel conditions. In particular, to support D2D communications among users to lower the energy wasted and improve the convergence time of the FL process, the use of an echo state network, running in local at each user site, is discussed. Numerical results are provided to highlight the suitability of the considered approach for applications foreseen for the forthcoming 6G networks.