Page 166 - Proceedings of the 2018 ITU Kaleidoscope

S2.4 Towards Cognitive Autonomous Network in 5G
Stephen S. Mwanje and Christian Mannweiler (Nokia Bell Labs, Germany)

Cell densification and addition of new Radio Access Technologies have been the solutions of
choice for improving area-spectral efficiency to serve the ever-growing traffic demand. Both
solutions, however, increase the cost and complexity of network operations for which the agreed
solution is increased automation. Cognitive Autonomous Networks (CAN) will therefore use
Artificial Intelligence and Machine Learning (ML) to maximize the value of automation. This
paper develops the models for cognitive automation and proposes a CAN design that addresses
the requirements for 5G and future networks. We then illustrate the benefit of this approach by
evaluating ML models that learn a network's response to different mobility states and
configurations.

Session 3: Machine Learning in Telecommunication Networks - II
Invited Paper - Machine Learning Opportunities in Cloud Computing Data Center Management
S3.1
for 5G Services
Fabio López-Pires (Itaipu Technological Park, Paraguay); Benjamín Barán (National University

of the East, Paraguay)
Emerging paradigms associated with cloud computing operations are considered to serve as a basis
for integrating 5G components and protocols. In the context of resource management for cloud
computing data centers, several research challenges could be addressed through state-of-the-art
machine learning techniques. This paper presents identified opportunities on improving critical
resource management decisions, analyzing the potential of applying machine learning to solve
these relevant problems, mainly in two-phase optimization schemes for virtual machine placement
(VMP). Potential directions for future research are also presented.

S3.2 Consideration on Automation of 5G Network Slicing with Machine Learning
Ved P. Kafle, Yusuke Fukushima; Pedro Martinez-Julia and Takaya Miyazawa (National Institute
of Information and Communications Technology, Japan)

Machine learning has the capability to provide simpler solutions to complex problems by
analyzing a huge volume of data in a short time, learning for adapting its functionality to
dynamically changing environments, and predicting near future events with reasonably good
accuracy. The 5G communication networks are getting complex due to emergence of
unprecedentedly huge number of new connected devices and new types of services. Moreover, the
requirements of creating virtual networkslices suitable to provide optimal services for diverse users
and applications are posing challenges to the efficient management of network resources,
processing information about a huge volume of traffic, staying robust against all potential security
threats, and adaptively adjustment of network functionality for time-varying workload. In this
paper, we introduce about the envisioned 5G network slicing and elaborate the necessity of
automation of network functions for the design, construction, deployment, operation, control and
management of network slices. We then revisit the machine learning techniques that can be applied
for the automation of network functions. We also discuss the status of artificial intelligence and
machine learning related activities being progressed in standards development organizations and
industrial forums.

– 150 –