where both xik and x jk are sets of values and Δ is the symmetric difference set operator.
This measure counts how many categories there are to which one of the products
at hand belongs and the other not.
As mentioned in Section 17.2.3, the aim of MDS is to map a dissimilarity matrix
Δ (having elements δi j and in our approach computed using (17.1)) as good as
possible to distances between points in a low dimensional Euclidean space. This
objective can be formalized by minimizing the raw Stress function [30]