Elastic Federated Learning (FL) is introduced to address challenges in implementing FL over 5G-and-beyond wireless networks in IoT environments.
Elastic FL leverages the Open RAN (O-RAN) systems and employs a novel resource provisioning methodology for executing FL-services.
It incorporates a multi-time-scale FL management system with three dedicated network control functionalities: non-real-time system descriptor, near-RT FL controller, and FL MAC scheduler.
The prototype of Elastic FL demonstrates its potential in improving the performance of FL-services.