Table 4.
Advantages and limitations of some selected graphs distance measures.
Characteristics
Spatial locationDifferent sizeComputational costStructural differenceAvailable Code
Methods SimiNet (Mheich et al., 2018) − ++ − https://github.com/amheich/SimiNet
D-measure (Schieber et al., 2017) − − + (not for sparse graph) https://github.com/tischieber/Quantifying-Network-Structural-Dissimilarities
DeltaCon (Koutra et al., 2013) − − https://web.eecs.umich.edu/∼dkoutra/CODE/deltacon.zip
Kernel methods (Borgwardt et al., 2005; Shervashidze et al., 2011) − − − https://github.com/BorgwardtLab/GraphKernels
Characteristics
Spatial locationDifferent sizeComputational costStructural differenceAvailable Code
Methods SimiNet (Mheich et al., 2018) − ++ − https://github.com/amheich/SimiNet
D-measure (Schieber et al., 2017) − − + (not for sparse graph) https://github.com/tischieber/Quantifying-Network-Structural-Dissimilarities
DeltaCon (Koutra et al., 2013) − − https://web.eecs.umich.edu/∼dkoutra/CODE/deltacon.zip
Kernel methods (Borgwardt et al., 2005; Shervashidze et al., 2011) − − − https://github.com/BorgwardtLab/GraphKernels

Note. Note that “−” indicates a characteristic that is not integrated in the similarity score of the method; “+” a characteristic that is integrated in the methods; “− −” for worst computational time; and “++” for very good computational time. Spatial location = physical location of nodes; Different size = graphs with different number of nodes; Computational cost = algorithm running time; Structural difference = detection of difference between node’s links in two graphs.

Close Modal