Conclusions
Our main results may be summarized as follows. The statistical problem of multialternative classification under constraints is stated as a problem of linear programming in functional space.
Necessary and sufficient conditions are obtained for the existence of an optimal decision, in the form of a functional analog of the Kuhn-Tucker theorem, yielding a system of equations and inequalities for finding the optimal decision rules.
The structure of the optimal rules is obtained for two types of functional and six types of constraint, which are given a meaningful treatment in the context of medical diagnosis problems.
Sufficient conditions for the optimal rules to be unrandomized are obtained in some particular cases.
Similar content being viewed by others
Literature Cited
S. Saks, Theory of Statistical Inferences [Russian translation], Mir, Moscow (1975).
M. Kendall and A. Stuart, Statistical Inference and Statistical Relationship, Hafner (1966).
G. E. Antonov, L. D. Chireikin, and D. Ya. Shurygin, “Automatic syndromic ECG diagnosis exemplified by changes in auricle of the heart,” in: Automatic Analysis of Electrocardiograms [in Russian], Meditsina, Moscow (1977).
A. S. Fomin, “Multialternative procedure for choosing decision in conditions of indeterminacy,” Izv. Akad. Nauk SSSR, Tekh. Kibern., No. 5, 155–160 (1976).
A. S. Fomin, “Construction of multialternative procedure for choosing decision in conditions of indeterminacy,” Izv. Akad. Nauk SSSR, Tekhn. Kibernetika, No. 6, 166–169 (1976).
W. Zangwill, Nonlinear Programming, Prentice-Hall, Englewood Cliffs, N. J. (1969).
L. Hurwitz, “Programming in linear spaces,” in: Studies in Linear and Nonlinear Programming [Russian translation], IL, Moscow (1962).
A. N. Kolmogorov and S. V. Fomin, Elements of Theory of Functions and Functional Analysis [in Russian], Nauka, Moscow (1968).
E. G. Gol'shtein, Theory of Duality in Mathematical Programming and Its Application [in Russian], Nauka, Moscow (1971).
Additional information
Translated from Kibernetika, No. 2, pp. 83–91, March–April, 1981.
Rights and permissions
About this article
Cite this article
Kul'chitskii, O.Y., Sloushch, A.V. & Sokolov, Y.V. On statistical classification problems with error constraints. Cybern Syst Anal 17, 248–258 (1981). https://doi.org/10.1007/BF01069642
Received:
Issue Date:
DOI: https://doi.org/10.1007/BF01069642