Advances in Pattern Recognition

Comparisons made in two studies of 21 methods for finding prototypes upon which to base the nearest prototype classifier are discussed. The criteria used to compare the methods are by whether they: (i) select or extract point prototypes; (ii) employ preor post-supervision; and (iii) specify the number of prototypes a priori, or obtain this number “automatically”. Numerical experiments with 5 data sets suggest that pre-supervised, extraction methods offer a better chance for success to the casual user than postsupervised, selection schemes. Our calculations also suggest that methods which find the "best" number of prototypes "automatically" are not superior to user specification of this parameter.