FG ML5G – Unified architecture for machine learning in 5G and future networks

This document defines a unified architecture for Machine Learning in Fifth Generation and future networks. A comprehensive set of architectural requirements is presented, which leads to specific architecture constructs needed to satisfy these requirements. Based on these constructs, a logical ML pipeline (along with the requirements derived and its realizations in various types of architectures) is discussed. Finally, key architectural issues facing the integration of such an ML pipeline into continuously evolving future networks are also listed.