A general structure for classification

The problem of classifying signals is considered. Typically, a finite number m of pairwise disjoint sets C/sub 1/,. . .,C/sub m/ of signals is given, and the aim is to synthesize a system that maps the elements of each C/sub j/ into a real number a/sub j/ so that the numbers a/sub 1/,. . .,a/sub m/ are distinct. A structure that can perform this classification (for any prescribed distinct a/sub 1/,. . .,a/sub m/), assuming only that the C/sub j/ are compact subsets of a real normed linear space, is presented.<
>