Induction of node label controlled graph grammar rules

Algorithms for inducing graph grammars from sets of graphs have been proposed before. An important class of such algorithms are those based on the Subdue graph mining system. But the rules learned by Subdue and its derivatives do not fit easily in any of the well-studied graph grammars formalisms. In this paper, we discuss how Subdue-like algorithms could be made to work in the context of NLC grammars, an important class of node replacement graph grammars. More specifically, we show how, given a set of occurrences of a subgraph, an NLC grammar rule can be induced such that the given occurrences could have been generated by it.

[1]  Lawrence B. Holder,et al.  Substructure Discovery Using Minimum Description Length and Background Knowledge , 1993, J. Artif. Intell. Res..

[2]  Joost Engelfriet,et al.  Graph Grammars Based on Node Rewriting: An Introduction to NLC Graph Grammars , 1990, Graph-Grammars and Their Application to Computer Science.

[3]  George Karypis,et al.  Finding Frequent Patterns in a Large Sparse Graph* , 2005, Data Mining and Knowledge Discovery.

[4]  Lawrence B. Holder,et al.  Mdl-based context-free graph grammar induction and applications , 2004, Int. J. Artif. Intell. Tools.