2 Neural methods from similarity-based perspective
The classification problem (the same reasoning may also be applied to regression and pattern
completion problems) is stated as follows: given a set of Nt class-labeled training vectors
{Rj
,C(Rj
)}, j = 1..Nt, where C(Rj
) is the class of Rj
, and given a vector X of an unknown
class, use the information provided in the similarity measure D(X,R j
) to estimate the
probability of classification p(Ci|X;M), where M describes the classification model used