Модели распознающих операторов, основанных на радиальных функциях
Аннотация
В статье рассмотрены вопросы, связанные с построением модели распознающих операторов, ориентированных на классификацию объектов в условиях большой размерности признакового пространства. В качестве исходной модели рассмотрена модель распознающих операторов, основанных на радиальных функциях. Основная идея предлагаемой модели состоит в формировании независимых подмножеств взаимосвязанных признаков, выделении набора предпочтительных пар репрезентативных признаков. Отличительная особенность предлагаемой модели алгоритмов заключается в определении подходящего набора двумерных функций расстояния при построении экстремального алгоритма распознавания. Целью данной статьи является разработка модели распознающих операторов, основанных на радиальных функциях, с использованием метода группового учета аргументов. Для проверки работоспособности предложенной модели распознающих алгоритмов проведены экспериментальные исследования при решении ряда задач. Анализ полученных результатов показывает, что рассмотренные модели распознающих операторов эффективно используются в тех случаях, когда между признаками объектов существует некоторая зависимость. При слабом выражении этой зависимости используется классическая модель распознающих операторов. Основным преимуществом предлагаемых распознающих операторов является улучшение точности и значительное сокращение объёма вычислительных операций при распознавании неизвестных объектов, что позволяет применить их при построении распознающих систем, работающих в режиме реального времени.
The problems of constructing of a model of recognition operators, oriented to the classification of objects in conditions of large dimensionality of the feature space are considered in this paper. As the initial model, the model of recognition operators, based on radial functions, are considered. The main idea of the proposed model consists in the formation of independent subsets of correlated features, the selection of a set of preferred pairs of representative features. A distinctive feature of the proposed model of algorithms is the determination of a suitable set of two-dimensional distance functions in the construction of an extreme recognition algorithm. The purpose of this article is to develop a model of recognition algorithms based on radial functions, using the method of group accounting of arguments. The subjects of the study are models of algorithms based on two-dimensional distance functions. In scientific terms, the results of this work together represent a new solution to the scientific problem related to the reliability of algorithms based on radial functions. The practical significance of the results lies in the fact that the developed algorithms and programs can be applied in medical and technical diagnostics, geological prediction, biometric identification and other areas where the problem of classification of objects specified in the large dimensional feature space. To test the efficiency of the proposed model of recognition algorithms, experimental studies were conducted to solve a number of model problems. The analysis of the obtained results shows that the considered models of algorithms are effectively used in those cases when there is a certain correlation between the features of the objects. When this correlation is weakly expressed, the classical model of recognition algorithms is used. The main advantage of the proposed recognition algorithms is the improvement of accuracy and a significant reduction in the amount of computational operations for the recognition of unknown objects, which makes it possible to apply them in the construction of real time operating recognition.
The problems of constructing of a model of recognition operators, oriented to the classification of objects in conditions of large dimensionality of the feature space are considered in this paper. As the initial model, the model of recognition operators, based on radial functions, are considered. The main idea of the proposed model consists in the formation of independent subsets of correlated features, the selection of a set of preferred pairs of representative features. A distinctive feature of the proposed model of algorithms is the determination of a suitable set of two-dimensional distance functions in the construction of an extreme recognition algorithm. The purpose of this article is to develop a model of recognition algorithms based on radial functions, using the method of group accounting of arguments. The subjects of the study are models of algorithms based on two-dimensional distance functions. In scientific terms, the results of this work together represent a new solution to the scientific problem related to the reliability of algorithms based on radial functions. The practical significance of the results lies in the fact that the developed algorithms and programs can be applied in medical and technical diagnostics, geological prediction, biometric identification and other areas where the problem of classification of objects specified in the large dimensional feature space. To test the efficiency of the proposed model of recognition algorithms, experimental studies were conducted to solve a number of model problems. The analysis of the obtained results shows that the considered models of algorithms are effectively used in those cases when there is a certain correlation between the features of the objects. When this correlation is weakly expressed, the classical model of recognition algorithms is used. The main advantage of the proposed recognition algorithms is the improvement of accuracy and a significant reduction in the amount of computational operations for the recognition of unknown objects, which makes it possible to apply them in the construction of real time operating recognition.