A Similarity Measure Between Patterns with Nonindependent Attributes
Tetsuro Ito, Yoshifumi Kodama, Junichi Toyoda · IEEE Transactions on Pattern Analysis and Machine Intelligence · 1984
A generalized version of a set-theoretical measure for obtaining similarities between patterns with nonindependent attributes is presented. The dependence here is given by the pairwise correlation. Since the proposed measure needs no assumption of attribute independence, the resulting similarity values can reflect directly the relationships between the attributes.