• similarity
• graph algorithms
• graph
• attributed networks
• social network analysis

Attributed networks store multiple values in their nodes, whose informative content is not always related to the structural content provided by the networks’ links.
While the literature has extensively studied many methods to perform clustering, community detection, and graph and pattern mining in these networks, less effort has been devoted to investigate similarity as a main goal (except for the similarity inside clusters and communities).
In this thesis, we exploit the attributes of the nodes found along in fixed length paths to define signatures that enrich the attribute content.
We provide experiments to motivate the significance of signatures on real-world networks with up to 30 million edges, fast algorithms to build sampled signatures for large graphs and discuss some applications such as node similarity, role similarity, and link prediction, for which our (sampled) signatures provide interesting experimental figures.