MACHINE LEARNING TECHNIQUES FOR 5G AND BEYOND

Network embedding assigns network nodes to low dimensional representations, which effectively preserves the network topology. Recent years have seen a significant degree of advancement towards this new paradigm for network research. In this work, we focus on categorizing, evaluating, and suggesting future paths for the study of network embedding techniques. We start by briefly outlining network embedding's goal. In the context of cognitive radio, we discuss network embedding and its relation to classical graph embedding techniques. Subsequently, we provide a comprehensive and systematic description of a wide range of network embedding strategies, such as sophisticated information preservation techniques, side information containing techniques, and approaches that maintain structure and attributes. Furthermore, a variety of network embedding assessment techniques are investigated, along with a few useful web resources including network data sets and software. In the final section, we discuss the fundamentals of applying these network embedding techniques to build a functional system.