Unlike centralized machine learning, FL involves training models across distributed data sources without moving the data. This introduces specific challenges around governance, interoperability, coordination, and secure infrastructure.
Editor
KU Leuven / University of UHasselt