Many flexible methods for graph dissimilarity computation are based on the concept of edit distance. A recently developed approximation framework allows one to compute graph edit distances substantially faster than traditional methods. Yet, this novel procedure considers the local edge structure only during the primary optimization process. Hence, the speed up is at the expense of an overestimation of the true graph edit distances in general. The present paper introduces an extension of this approximation framework. Regarding the node assignment from the original approximation as a starting point, we implement a search procedure based on a genetic algorithm in order to improve the approximation quality. In an experimental evaluation on three real world data sets a substantial gain of distance accuracy is empirically verified. © 2014 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Riesen, K., Fischer, A., & Bunke, H. (2014). Improving approximate graph edit distance using genetic algorithms. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8621 LNCS, pp. 63–72). Springer Verlag. https://doi.org/10.1007/978-3-662-44415-3_7
Mendeley helps you to discover research relevant for your work.