Neural network modeling for student competency development

Kostiuk, Yuliia та Rzaieva, Svitlana та Skladannyi, Pavlo та Sokolov, Volodymyr (2026) Neural network modeling for student competency development The Ninth International Workshop on Computer Modeling and Intelligent Systems (CMIS-2026) (4220). с. 252-263. ISSN 1613-0073

Текст Y_Kostiuk_S_Rzaieva_P_Skladannyi_V_Sokolov_CMIS_2026_FITM.pdf Download (744kB)

Анотація

The rapid growth of data availability and advances in computing power have positioned neural network modeling and machine learning as essential tools for developing key competencies among higher education students. This paper investigates the application of neurostructural modeling—a generalized extension of traditional feed-forward artificial neural networks—to analyze experimental and observational data, with the primary aim of enhancing students’ data analysis, predictive modeling, and decision-making skills in modern educational contexts. A unified theoretical framework is proposed that treats neural network models as compositions of linear and nonlinear structures, incorporating nonclassical activation functions (periodic and parameterized forms) and generalized neuron-like elements with flexible connectivity across layers. Special attention is given to constructive (incremental) algorithms for building neurostructural models that guarantee monotonic reduction of the training error. Training is formulated as a nonlinear least-squares problem and addressed through a class of efficient numerical methods based on linear-nonlinear weight decomposition, pseudo-inversion (including block pseudorotation via Klines’s formula), and Gauss-Newton-like updates with backward-propagation-style Jacobian computation. The developed approach is implemented in a software suite for data storage, extraction, neural model construction, training, and application, including modules for cluster analysis (Kohonen networks), optimal control of dynamic systems with long-term prediction horizons, and analytical processing of large datasets. Experimental validation demonstrates improved computational stability, faster convergence in many cases compared to classical backpropagation, and practical utility for modeling complex systems. The results confirm that integrating neurostructural modeling techniques into higher education curricula significantly strengthens students’ competencies in machine learning, datadriven modeling, and adaptive system analysis—skills increasingly demanded in contemporary professional environments. Directions for future work include integration with real-time adaptive learning platforms to support personalized competency development

Тип елементу :	Стаття
Ключові слова:	neural network model; feed-forward neural network; optimal control; higher education student
Типологія:	Статті у базах даних > Scopus (без квартилю)
Підрозділи:	Факультет інформаційних технологій та математики > Кафедра інформаційної та кібернетичної безпеки ім. професора Володимира Бурячка
Користувач, що депонує:	Павло Миколайович Складанний
Дата внесення:	12 Серп 2026 17:10
Останні зміни:	12 Серп 2026 17:10
URI:	https://elibrary.kubg.edu.ua/id/eprint/59144

Actions (login required)

Перегляд елементу