Relational models for generating labeled real-world graphs

Analyzing and understanding the structure of social networks and other real-world graphs has become a major area of research in the field of data mining. An important problem setting is the creation of realistic synthetic graphs that resemble realworld social networks. While a range of efficient algorithms for this task have been proposed, current methods solely take the network topology into account ignoring any node labels. We propose a probabilistic approach to synthetic graph generation with node labels, building on concepts from relational learning.