Frequent Submap Discovery

Combinatorial maps are nice data structures for modeling the topology of nD objects subdivided in cells (e.g., vertices, edges, faces, volumes, ...) by means of incidence and adjacency relationships between these cells. In particular, they can be used to model the topology of plane graphs. In this paper, we describe an algorithm, called mSpan, for extracting patterns which occur frequently in a database of maps. We experimentally compare mSpan with gSpan on a synthetic database of randomly generated 2D and 3D maps. We show that gSpan does not extract the same patterns, as it only considers adjacency relationships between cells. We also show that mSpan exhibits nicer scale-up properties when increasing map sizes or when decreasing frequency.