Krykun V. Method and software tools for interpreting machine learning models of nonlinear dynamic objects

Українська версія

Thesis for the degree of Doctor of Philosophy (PhD)

State registration number

0824U002664

Thesis Registration Form

0824U002664.pdf

Applicant for

Specialization

  • 121 - Інженерія програмного забезпечення

01-07-2024

Specialized Academic Board

5770

Odesa Polytechnic National University

Essay

The dissertation is devoted to solving an urgent scientific and practical problem, which is to create a method for interpreting machine learning models of nonlinear dynamic objects with continuous characteristics and its application in the form of software and algorithmic identification tools as an integral part of intelligent systems. The purpose of the study is to improve the accuracy of surrogate models by developing a method for interpreting machine learning models of nonlinear dynamic objects using nonparametric dynamic models based on integral-power Volterra polynomials and implementing the proposed method in the form of software and algorithmic tools for identifying continuous objects as part of intelligent systems. To achieve the research purpose, the following tasks were set and solved: the analysis of existing methods for interpreting machine learning models of continuous nonlinear dynamic objects is carried out, the basic problems of interpreting neural network models are identified; the choice of research direction in the field of building surrogate models in the form of integral nonparametric dynamic models based on integral-power Volterra polynomials for the analytical description of nonlinear dynamic objects is substantiated; the use of nonparametric dynamic models based on Volterra integral polynomials as surrogate models for the interpretation of neural networks with time delays was proposed; an analytical connection between neural networks with time delays and nonparametric dynamic models based on Volterra integral polynomials was established to improve the accuracy of surrogate models of continuous nonlinear dynamic objects; a method for interpreting neural networks with time delays was developed by constructing surrogate models in the form of nonparametric dynamic models based on Volterra integral-power polynomials to describe nonlinear dynamic objects with acceptable modeling accuracy; an information technology for interpreting machine learning models of nonlinear dynamic objects based on the use of a neural network with time delays to identify a nonlinear dynamic model of an object based on input-output experiment data and to build surrogate models in the form of nonparametric dynamic models based on integral-power Volterra polynomials with acceptable modeling accuracy for interpreting the neural network was developed; computer modeling tools for constructing neural network models of nonlinear dynamic objects and surrogate models for their interpretation in continuous object identification systems was developed; the developed method and tools to solve applied problems of continuous object identification was applied.

Read more

Research papers

Krykun V. Improving the accuracy of the neural network models interpretation of nonlinear dynamic objects. Математичне та комп'ютерне моделювання. Серія: Технічні науки. 2023. Вип. 24. C. 45–55. DOI: 10.32626/2308-5916.2023-24.45-55.

Mathematical models of software quality assurance for interpretation of dynamic neural networks / O. Fomin, V. Krykun, A. Orlov et al. Вчені записки ТНУ імені В.І. Вернадського. Серія: Технічні науки. 2023. Том 34 (73), № 5. C. 250–256. DOI: 10.32782/2663-5941/2023.5/39.

Models of dynamic objects with significant nonlinearity based on time-delay neural networks / O. Fomin, V. Speranskyy, V. Krykun et al. Вісник черкаського державного технологічного університету. Технічні науки. 2023. № 3. С. 97–112. DOI: 10.24025/2306-4412.3.2023.288284.

Fomin O.O., Krykun, V.A. Assessment of the Quality of Neural Network Models Based on a Multifactorial Information Criterion. Вісник сучасних інформаційних технологій. 2024. Том 7, № 1. С. 13–23. DOI: 10.15276/hait.07.2024.1.

Interpretation of Dynamic Models Based on Neural Networks in the Form of Integral-Power Series / O. Fomin et al. ; in: Arsenyeva, O., Romanova, T., Sukhonos, M., Tsegelnyk, Y. (eds). Smart Technologies in Urban Engineering. STUE 2022. Lecture Notes in Networks and Systems. Springer, Cham, 2022. Vol. 536. P. 258–265. DOI: 10.1007/978-3-031-20141-7_24.

Interpretation Method for Dynamic States Neural Network Models / S. Polozhaenko et al. IEEE 3rd International Conference on System Analysis & Intelligent Computing (SAIC). Kyiv, Ukraine, 2022. P. 1–5. DOI: 10.1109/SAIC57818.2022.9923008.

Крикун В. А., Фомін О. О. Інтерпретація динамічних моделей у вигляді нейронних мереж з часовими затримками. Матеріали Дванадцятої Міжнародної наукової конференції студентів та молодих учених «Сучасні інформаційні технології – 2022». Одеса, Україна, 2022. С. 140-141.

Use of dynamic neural networks for modeling nonlinear objects with significant nonlinearity / S. Polozhaenko et al. Збірник тез IV Міжнародної науково-практичної Інтернет-конференції "Математика та інформатика в науці й освіті: виклики сучасності". Вінниця, Україна, 2023. C. 121-124.

Крикун В. А., Фомін О. О. Нелінійне моделювання об'єктів на основі динамічних нейронних мереж. Матеріали Тринадцятої Міжнародної наукової конференції студентів та молодих учених «Сучасні інформаційні технології – 2023». Одеса, Україна, 2023. С. 153-155.

Use of Dynamic Neural Networks for Modeling Nonlinear Objects with Significant Nonlinearity / O. Fomin et al. 18th Conference on Computer Science and Intelligence Systems. Warsaw, Poland, 2023. P. 97–102. DOI: 10.15439/2023F3874.

Modeling of the agricultural crops development using satellite imagery / O. Fomin et al. Біоінтенсивні та SMART-технології у тваринництві: матеріали II Міжнародної науково-практичної конференції науково-педагогічних працівників та молодих науковців. Одеса, Україна, 2023. C. 10–14.

Крикун В. А. Математична модель оцінки якості програмного забезпечення інтерпретації моделей машинного навчання. Комп’ютеризовані системи та програмні технології. 2023. № 1. С. 7–11.

Read more

Files

autoreferat-abstract_krikun.pdf

dis_krikun.pdf

Similar theses

0824U002621

Vadym ZABCHUK

2. Methods and software tools for modeling processes in biogas plants based on interval data analysis

0824U002601

Oleksandra D. Syzonenko

Development of technology and software tools for detecting and recognizing objects in real time

0824U002554

Bohdan V. Molodets

Development of technologies and software for operational air quality monitoring

0824U002508

Vladyslav Antoniuk

Development of a cross-platform system for automating processes and interacting with computerized devices

0824U002458

Oleksandr Lemeshchuk

Development of the University Quality Management System by means of Services