Graph Indexing and Retrieval Based on Median Graphs

Metric indexing is used to organize large databases of Attributed Graphs and define fast queries on these sets. We present a proposal where indexing is based on an m-tree in which routing nodes of the m-tree act as prototypes of the sub-cluster of Attributed Graphs that routing nodes represent. In the classical schemes, the information kept in the tree nodes of the subcluster is a selected Attributed Graph from the sub-set. Depending on the subcluster and the application, it is difficult to select a good representative of the sub-clusters. To that aim, we propose to use Median Graphs as the main information kept in the routing nodes of the m-tree. Experimental validation shows that in the database queries, the decrease of the nodes explored in the mtree while using a Generalized Median Graph is about 20% respect using a selected Attributed Graph.