CONSTRUCTION OF CONFIDENCE BOUNDARIES FOR THE DECISION FUNCTION IN A TWO-ALTERNATIVE PROBLEM OF PATTERN RECOGNITION
A. V. Lapko1,2, V. A. Lapko1,2
1Institute of Computational Modeling, Siberian Branch, Russian Academy of Sciences, Akademgorodok 50, building 44, Krasnoyarsk, 660036 Russia
2Reshetnev Siberian State Aerospace University, pr. im. Gazety “Krasnoyarskii rabochii” 31, Krasnoyarsk, 660014 Russia
Keywords: распознавание образов, решающая функция, непараметрическая оценка, доверительное оценивание, правило Хайнкольда - Гаеде, pattern recognition, decision function, nonparametric estimate, confidence estimate, Heinhold-Gaede rule
Abstract
A nonparametric estimate of the decision function in a two-alternative problem of pattern recognition is considered. The principle of expansion of the learning sample and the analysis of the probabilistic characteristics of the obtained sets of random variables are used to synthesize this estimate. On this basis, a technique of construction of confidence boundaries for the Bayesian equation of the separating surface is derived. The efficiency of this technique is proved by results of computational experiments.
|
