INFORMATION-EXTREME MACHINE TRAINING SYSTEM OF FUNCTIONAL DIAGNOSIS SYSTEM WITH HIERARCHICAL DATA STRUCTURE

Context. The problem of information-extreme machine learning the functional diagnosis system is considered by example recognizing technical state a laser printer typical defects printed material. object research process hierarchical an electromechanical device. Objective. main objective to improve efficiency during diagnostics retraining using automatically forming new data structure for expanded alphabet recognition classes. Method. A method based on material proposed. was developed with approach modeling cognitive processes natural intelligence, which makes it possible give diagnostic properties adaptability under arbitrary initial conditions formation images printing and flexibility due increase in power principle maximizing amount information learning. as iterative procedure optimizing parameters functioning according criterion. As criterion parameters, modified Kullback’s measure considered, exact characteristics classification solutions. According proposed categorical model, algorithm has been form binary decomposition tree. use such split large number classes into pairs nearest neighbors, optimization carried out linear required depth. Results. Information, algorithmic software developed. influence investigated. Conclusions. results physical have confirmed can be recommended practical use. prospect increasing information-extremal depth additional system’s functions, including input training matrix.