Attributed Graph Pattern Set Selection Under a Distance Constraint.

Pattern mining exhaustively enumerates patterns that occur in some structured dataset and each satisfy some constraints. To avoid redundancy and reduce the set of patterns resulting from the enumeration, it is necessary to go beyond the individual selection of patterns and select a pattern subset which, as a whole, contains relevant and non redundant information. This is particularly useful when enumerating bi-patterns, which represent pairs of attribute patterns describing for instance subnetworks in two-mode attributed networks. We present and experiment a general greedy algorithm performing pattern set selection on attributed graphs.