Upper Bounding Graph Edit Distance Based on Rings and Machine Learning

The graph edit distance (GED) is a flexible measure which widely used for inexact matching. Since its exact computation [Formula: see text]-hard, heuristics are in practice. A popular approach to obtain upper bounds GED via transformations the linear sum assignment problem with error-correction (LSAPE). Typically, local structures and distances between them employed carrying out this transformation, but recently also machine learning techniques have been used. In paper, we formally define unifying framework LSAPE-GED from LSAPE. We introduce rings, new kind of designed graphs where most information resides topology rather than node labels. Furthermore, propose two ring-based RING RING-ML, instantiate using traditional learning-based transforming LSAPE, respectively. Extensive experiments show that rings bounding significantly improves state art on datasets graphs’ topologies. This closes gap fast inaccurate LSAPE-based more accurate slower algorithms based search.