A novel software toolkit for graph edit distance computation

Abstract

Graph edit distance is one of the most flexible mechanisms for error-tolerant graph matching. Its key advantage is that edit distance is applicable to unconstrained attributed graphs and can be tailored to a wide variety of applications by means of specific edit cost functions. The computational complexity of graph edit distance, however, is exponential in the number of nodes, which makes it feasible for small graphs only. In recent years the authors of the present paper introduced several powerful approximations for fast suboptimal graph edit distance computation. The contribution of the present work is a self standing software tool integrating these suboptimal graph matching algorithms. It is about being made publicly available. The idea of this software tool is that the powerful and flexible algorithmic framework for graph edit distance computation can easily be adapted to specific problem domains via a versatile graphical user interface. The aim of the present paper is twofold. First, it reviews the implemented approximation methods and second, it thoroughly describes the features and application of the novel graph matching software. © 2013 Springer-Verlag.