APPLICATION OF THE ARTIFICIAL NEURAL NETWORKS THEORY IN PROBLEMS OF APPLIED PEDAGOGY OF HIGHER EDUCATION INSTITUTIONS

Abstract

The paper deals with the problem of social system analysis, which consists of a large number of students and teachers of a higher educational institution for the creation of effective educational pairs of a teacher-student. As an example, we used the elementary structures of Lviv Polytechnic National University - graduating departments. Elements of artificial intelligence theory based on artificial neural networks were used to form the mentioned training pairs. Two types of research are analyzed. The first type of research was to model effective training pairs of a student-teacher, and the second type was to isolate these effective pairs from a given sample size, that is, the clustering problem was solved. For the first type of tasks, we used the artificial neural network machine trained with the supervisor, in particular, the wellknown backpropagation method, implemented with the Matlab for Windows programming language. For the second type of tasks, the device used artificial neural networks trained without a supervisor, in particular, the famous method WTA (Winner Takes All - the winner takes everything), also implemented in the language of Matlab for Windows.
It is also considered the possibility of improving our modeling and clustering methods by expanding the size of training vectors for supervisor networks, as well as expanding the coefficient matrix for clustering problems. The possibility of applying our methods to the problems of applied psychology and pedagogy with respect to such high sampling social systems is substantiated. The analysis of the conducted researches is carried out, as well as the development prospects of the latter are presented. The hypothesis about the possibility of combining modeling and clustering problems into a single one by expanding the order of the input vector columns and the coefficient matrix, that is, in mathematical interpretations of the implementation of a higher rank tensor, was hypothesized.