Sensitivity analysis for structural matching

The aim of this paper is to explore the sensitivity of relational matching to attribute and structural information. Broadly speaking there are two main aspects to this study. First, we determine the relative importance of attribute and structural information in matching noise corrupted graphs. The second aspect of our analysis concerns the nature of the relational structures used in matching. Here we compare the matching results obtained using four different graph structures, namely the Delaunay graph, the N-nearest neighbour graph, the Gabriel graph and the relative neighbourhood graph. Our results are presented as noise sensitivity curves. The main conclusion of the study is that attributes are essential when the fractional corruption exceeds 20% and that the Delaunay graph has optimal noise robustness