Machine Learning Methods with Node-Level Features

Abstract

In this paper we describe several methods for obtaining node features. They can also be grouped based on structure-based features, which again, do not require degrees, such as the simplest one, and which count edges, clusters, and triangles, as well as significance features like node degree and various centrality measures. A generalization that counts additional structures in which a particular node of interest participates is the Graphlet Degree Vector (GDV).