Abstract

To support heterogeneous network services with strict requirements for various performance characteristics, 5G networks leverage network softwarization and virtualization paradigms such as software defined networking (SDN), network functions virtualization (NFV), and network slicing. The resulting architecture features a high degree of complexity with numerous interfaces to achieve additional goals like programmability, automation, service-awareness, and resource efficiency. Recent advances in the domain of Machine Learning (ML) and Artificial Intelligence (AI) make mechanisms from these domains promising candidates to cope with the complexity regarding the management and orchestration (MANO) of physical and virtual network resources as well as to translate technical Quality of Service (QoS) parameters into user-centric Quality of Experience (QoE) information. In this chapter, we discuss where and how ML and AI techniques can be applied in the 5G ecosystem. We cover three representative case studies focusing on QoE assessment, deployment of Virtualized Network Functions (VNFs), and slice management. Furthermore, we point out general and use case-specific requirements and challenges, and derive guidelines for operators who plan to deploy such mechanisms in their network.