Abstract

The work of Ioannidis et al. 2018 introduces a family of distances betweentwo graphs that provides tractable graph alignment strategies. Importantly, thealignment scores produced by this family satisfy the properties of metrics,which is very useful in several learning tasks. In this paper, we generalizethis work to compare n graphs by introducing a family of distances, which is ann-metric, i.e., an extension of a metric to n elements that includes ageneralization of the triangle inequality. Our new family of distances,includes the ones in the work of Ioannidis et al. 2018 as a special case, andcan produce tractable alignments between multiple graphs.

 

Quick Read (beta)