Statistical approach to building a model of recognition operators under conditions of high dimensionality of a feature space

Abstract

Listen

Translate article

The task of constructing recognition algorithms (RAs) focused on the classification of objects under conditions of high dimensionality of a feature space is considered. As an initial model, a model of RAs based on the calculation of estimates is considered. A distinctive feature of the proposed model is the formation of subsets of interrelated features (SIF) and the selection of a set of representative features (RFs) in the construction of RAs. Moreover, the proximity estimates between the object and the class under consideration are calculated on the basis of the Bayesian approach. The main advantage of the proposed operators is the allocation of preferred dependency models with the subsequent calculation of the assessment of the ownership of objects and ensuring a significant reduction in the number of computational operations when recognition unknown objects. This feature is very important for real-time recognition systems. To test the performance of the proposed model, experimental studies were carried out in solving a model problem. This model can be used in the compilation of various programs aimed at solving problems of forecasting and classifying objects defined in the space of features of large dimension.