摘要: This paper is specifically concerned with the problem of inferring from a finite set patterns classification an unknown pattern. A discussion general problems inherent in concept “learning” and “data reduction” are discussed standpoint measurement selection for pattern recognition problem. brief history existent work empirical Bayes compound sequential procedures will be presented. It felt that these basically non-Bayesian, despite their names, therefore especially suited to arising recognition. Finally, made some nonparametric approaches when only information on underlying distributions associated various categories which can obtained number samples.