Кафедра штучного інтелекту (ШІ)

Постійний URI для цієї колекції


Останні подання

Зараз показано 1 - 5 з 164
  • Документ
    Дослідження продуктивності провідних JAVASCRIPT-фреймворків
    (ХНУРЕ, 2023) Дранченко, С. А.
    This research focuses on the evaluation of the performance of leading JavaScript frameworks. The study delves into the analysis of various frameworks and their respective performance metrics, such as load time, rendering speed, file size, lines of code. The study aims to assist developers in selecting the most suitable JavaScript framework for their specific projects based on performance requirements.
  • Документ
    CLIPTraVeLGAN for Semantically Robust Unpaired Image Translation
    (2023) Bodyanskiy, Y.; Ryabova, N.; Lavrynenko, R.
    In this paper a novel approach for semantically robust unpaired image translation is presented. CLIPTraVeLGAN replaces the Siamese network in TraVeLGAN with a contrastively pretrained language-image model (CLIP) with frozen weights. This approach significantly simplifies the model selection and training process of TraVeLGAN, making it more robust and easier to use.
  • Документ
    Матрична еволюційна нейро-фаззи система для швидкого розпізнавання образів-зображень
    (ХНУРЕ, 2020) Чала, О. С.
    In this thesis the task of data stream fuzzy classification that is fed to processing in online mode wherein the characteristics of the data can be changed in time in an unpredictable way. For the solution the evolving fuzzyprobabilistic neural network for fast image processing is proposed. Distinctive features of the system are, firstly, the data are fed to the network in matrix form, secondly, in process of learning the architecture of neural network, activation function parameters including centroids and width parameters tune which significantly raise the accuracy
  • Документ
    Порівняльний аналіз нейромережевих методів в задачах класифікації об'єктів
    (ХНУРЕ, 2020) Косолапов, К. С.
    Neural Networks based approach to object classifying is considered. The main difficulties of the object classification problem are analyzed. Such classification problems as binary are considered, as well as the task of classifying objects of the same type, which are subclassed by their characteristic features. Different types of classic and deep neural networks are described. The advantages and disadvantages of neural networks and deep neural networks are described. An example of the second type task is the classification of apples by their varieties. It is supposed to split this type of object into 13 classes. The use of convolutional neural networks for the task of classifying apples by varieties is substantiated.
  • Документ
    Two-dimensional matrix neural networks as alternative to modern state-of-the art ann architectures
    (ХНУРЕ, 2020) Албасова, А. І.
    This work is devoted to the study of some issues related to the use of traditional vector-based neural networks to process non-vectorial inputs such as matrices. As traditional ANNs assume vector inputs, non-vectorial information must be previously converted to vector form. This process causes some problems, e.g. lost of information and decrease in learning speed. The purpose of this work is to point out the limits of traditional ANNs in handling nonvectorial data and consider matrix neural networks, which takes matrices directly as inputs.