Структурно-параметричний синтез гібридних нейронних мереж

Скорочена інформація про документ
dc.contributor.authorЧумаченко, Олена Іллівна
dc.date.accessioned2019-10-04T12:36:46Z
dc.date.available2019-10-04T12:36:46Z
dc.date.issued2019
dc.descriptionРоботу виконано в Національному технічному університеті України «Київський політехнічний інститут імені Ігоря Сікорського» Міністерства освіти і науки.uk
dc.description.abstractenThe necessity of the development of integrated, hybrid systems based on deep learning is substantiated. Such systems consist of various elements (components), united in the interests of achieving objectives set. In the thesis the actual scientific-applied problem, which has the important scientific and practical importance, is solved and it consists in the development of methods and algorithms for solving the problem of structural-parametric synthesis of deep learning hybrid neural networks (HNN). It is shown that the main problems of synthesis of HNN, at present, are: – absence of formal methods for choosing the type of neural networks (NN), adequate for the class of tasks to be solved; – insufficient work on the issues of automatic formation of the topology of NN, which does not allow to create NN of high accuracy and minimum complexity (minimum computational costs); – insufficient grounds for choosing optimization methods in the training procedure of NN, which leads to significant errors. In the course of the thesis work the methodology of structural-parametric synthesis of HNN; method of structural-parametric synthesis of modules of HNN; algorithm of structural-parametric synthesis of ensemble of modules of HNN; algorithm of structuralparametric synthesis of HNN of deep learning; methods of prediction based on the use of HNN of deep learning is developed. A new methodology for the synthesis of HNN is developed, which is differed by the fact that in the first stage the optimal base neural network is selected; in the second stage, as a result of solving the multicriteria optimization problem, it is modified; in the third stage, the problem of structural-parametric synthesis of modules is considered at the fourth stage the problem of structural and parametric synthesis of the ensemble is solved, which allows to improve the accuracy of the systems operation in their minimal complexity. The problem of optimal choice of basic neural network (BNN) topology is solved by using the method of selection. Numerous examples of optimal choice of BNN are given in the work. A new method for modifying the BNN is developed, which in opposite to the known ones, in order to increase the efficiency of the problem solution (increase of accuracy and decrease the complexity of the BNN), the parameters adjustment is carried out in two stages: at the first stage a hybrid multicriteria evolutionary algorithm is used, and at the second one – for more accurate determination of neurons number in the hidden layers, an adaptive algorithm of constructing and pruning is applied, the weight coefficients are specified by the gradient descent method. It is proved that evolutionary algorithms are not well adapted for solving problems with constraints and require some modification taking into account the specifics of the conditional optimization problem. To eliminate the identified shortcomings of genetic algorithms, it is proposed to "treat" (refine) nonmotivated points obtained after stopping the genetic algorithm. Due to the fact, that solutions in genetic algorithms are presentedas a vectot, consist of zeros and units, for the ‘treatment’ of uncommitted points, it is very convenient to use the Pareto local search algorithms in space of Boolean variables. To solve conditional multi-criteria optimization tasks, it is proposed to use a hybrid genetic algorithm. Based on the analysis carried out, in this work, it is proposed to synthesize the hybrid topology in the form of a parallel ensemble of NN modules with a layer of association. As a procedure for building an ensemble, the use of begging, which has advantages over others, is substantiated. To optimize the size of the ensemble, an algorithm for simplification was developed with the help of the complementary value method, which also takes into account the interaction of the classifiers with each other. The weigh coefficients of the association of modules in the ensemble were determined on the basis of the use of the method of dynamic averaging. A new hybrid algorithm of deep learning neural network topology formation has been developed, which in opposite to the known ones the parameters of the main network are determined by the sequential execution of each search iteration sequentially with each of the basic algorithms (swarm particles and genetic), the comparison of the found results and the use of the best found solutions of each algorithm, that allows to increase the accuracy and speed of network work under minimal complexity. The methodology of images processing on the basis of convolution neural networks and deep learning classifiers for non-formalized descriptors detecting is improved, it is determined: the optimal parameters of convolution neural networks with help of genetic algorithm; the training sample is formed as a result of system approach which includes: removal of noise in the image, image segmentation, selection of borders in the picture, formalization of the object descriptor, classification of the descriptor, which makes it possible to increase the accuracy of recognition. The problems that arise when solving prediction problems of time series with a large number of input variables are determined. The hybrid method of solving forecasting problems is developed, which is distinguished by the fact that it implements deep learning based on the use of a single-layer network with neurons of the type sigm_piecewise, constructed using the group method of data handling method, with the subsequent learning of the entire network as a whole by the method of reverse error propagation in order to find the global extremum, which increases the accuracy of prediction. The regularization prediction method is developed which in opposite to the known ones, it can be used in the case of heterogeneity of data and is based on the use of a soft clustering algorithm, in which as a surface model, separating clusters, it is used single layer NN with sigm_piecewise neurons and local NN, one for each cluster whose training are carried out only on examples from one cluster, that increases the accuracy of prediction.uk
dc.description.abstractruЗадача синтеза гибридных нейронных сетей во время решения задач (аппроксимации, классификации и прогнозирования) поставлена и решена в двух постановках: оптимального выбора известной топологии нейронной сети с дальнейшей модификацией и синтезом субоптимальной топологии с заданными критериями. В работе предложен алгоритм наращивания топологий гибридной искусственной нейронной сети путём сложения модулей и соединения в последовательные и параллельные ансамбли. Разработан алгоритм формирования топологии искусственной нейронной сети с использованием генетического алгоритма. Рассмотрено решение задачи структурно-параметрического синтеза гибридных нейронных сетей глубокого обучения. Приведены решения прикладных задач предложенными методами.uk
dc.description.abstractukДисертацію присвячено подальшому розвитку теорії розробки та дослідженню методів та алгоритмів штучного інтелекту на основі використання гібридних нейронних мереж. Задача синтезу гібридних нейронних мереж під розв’язання задач (апроксимації, класифікації та прогнозування) поставлена та розв’язана у двох постановках: оптимального вибору відомої топології нейронної мережі із подальшою модифікацією та синтезу субоптимальної топології за заданими критеріями. У роботі запропоновано алгоритм нарощування топологій гібридної ШНМ шляхом додавання модулів і з’єднання в послідовні і паралельні ансамблі. Розроблено алгоритм формування топології ШНМ з використанням генетичного алгоритму. Розглянуто розв’язання задачі структурно-параметричного синтезу гібридних нейронних мереж глибокого навчання. Наведено розв’язання прикладних задач запропонованими методами.uk
dc.format.page42 с.uk
dc.identifier.citationЧумаченко, О. І. Структурно-параметричний синтез гібридних нейронних мереж : автореф. дис. … д-ра техн. наук. : 05.13.23 – системи та засоби штучного інтелекту / Чумаченко Олена Іллівна. – Київ, 2019. – 42 с.uk
dc.identifier.urihttps://ela.kpi.ua/handle/123456789/29642
dc.language.isoukuk
dc.publisherКПІ ім. Ігоря Сікорськогоuk
dc.publisher.placeКиївuk
dc.subjectгібридні нейронні мережіuk
dc.subjectсистеми штучного інтелектуuk
dc.subjectобробка зображеньuk
dc.subjectсистеми медичної та технічної діагностикиuk
dc.subjectавтоматизовані системи керування дорожнім рухомuk
dc.subjectінформаційні пожежні системиuk
dc.subjecthybrid neural networksuk
dc.subjectartificial intelligence systemsuk
dc.subjectimage processinguk
dc.subjectsystems of medical and technical diagnosticsuk
dc.subjectautomated traffic control systemsuk
dc.subjectinformation fire systemsuk
dc.subjectгибридные нейронные сетиuk
dc.subjectсистемы искусственного интеллектаuk
dc.subjectобработка изображенийuk
dc.subjectсистемы медицинской и технической диагностикиuk
dc.subjectавтоматизированные системы управления дорожным движениемuk
dc.subjectинформационные пожарные системыuk
dc.subject.udc004.032.26:004.8](043.3/.5)uk
dc.titleСтруктурно-параметричний синтез гібридних нейронних мережuk
dc.typeThesisuk

Файли

Контейнер файлів
Зараз показуємо 1 - 1 з 1
Ескіз
Назва:
Chumachenko_aref.pdf
Розмір:
963.61 KB
Формат:
Adobe Portable Document Format
Опис:
Завантажити
Ліцензійна угода
Зараз показуємо 1 - 1 з 1
Ескіз недоступний
Назва:
license.txt
Розмір:
8.98 KB
Формат:
Item-specific license agreed upon to submission
Опис:
Завантажити

Зібрання

Автореферати (вільний доступ)