Graph and Heterogeneous Network Transformations

This chapter presents the key ideas related to embedding simple (homogeneous) and more elaborate (heterogeneous) graphs. It first outlines some of the key approaches for constructing embeddings based exclusively on graph topology. We then discuss recent advances that include feature information into the constructed graph embeddings. We also address data transformation methods applicable to graphs with different types of nodes and edges, referred to as heterogeneous graphs or heterogeneous information networks. The chapter is structured as follows. Section 5.1 presents approaches to embedding simple homogeneous graphs, starting with the popular DeepWalk and node2vec methods, followed by a selection of other random walk based graph embedding methods. Section 5.2 introduces heterogeneous information networks, containing nodes of different types, the most useful tasks applied to such networks, and selected approaches to embedding heterogeneous information networks. We present a method for propositionalizing text enriched heterogeneous information networks and a method for heterogeneous network decomposition in Sect. 5.3. Ontology transformations for semantic data mining are presented in Sect. 5.4. Selected techniques for embedding knowledge graphs are presented in Sect. 5.5. The chapter concludes by presenting selected methods implemented in Jupyter Python notebooks in Sect. 5.6.