Donetsk National Technical University
Please use this identifier to cite or link to this item:
https://ea.donntu.edu.ua/jspui/handle/123456789/30369| Title: | Дослідження алгоритмів навчання штучного інтелекту в комп’ютерних іграх |
| Other Titles: | RESEARCH OF ALGORITHMS OF LEARNING ARTIFICIAL INTELLIGENCE IN COMPUTER GAMES |
| Authors: | Костюкова, Н.С. Глинська, К. С. |
| Keywords: | нейронні мережі штучний інтелект ігровий додаток MLP CNN GAN Python |
| Issue Date: | Nov-2018 |
| Publisher: | Донецький національний технічний університет |
| Series/Report no.: | Інформатика, кібернетика та обчислювальна техніка;№2 (27)’ 2018 |
| Abstract: | Розглянуто можливості впровадження різних топологій й конфігурацій нейронних мереж в ігрових додатках для підсилення штучного інтелекту. Досліджено вплив параметрів нейронних мереж на їх здатність до навчання. Проведено порівняння декількох моделей нейронних мереж за критеріями точності, часу, перемог. |
| Description: | The work is devoted to different approaches to the implementation of the process of learning game artificial intelligence by means of neural networks in order to enhance the intelligence of computer agents and the quality of the game process. The object of research is algorithms and methods for the implementation of artificial intelligence. The subject of the study is neural networks as a way of machine learning that solves the problem of creating artificial intelligence in situations where traditional algorithms and manual programming of actions are not possible. The use of neural networks makes it possible to simulate artificial intelligence, capable of selforganization, learning and self-learning, adapting behaviour to new conditions, memorizing and analysing input data. Today, the operation of neural networks is a promising issue, as the effective use of technology in modern gaming applications suggests, but there are still many difficulties that need to be solved, and their use in gaming applications is poorly understood and limited by obstacles that prevent the use of all the benefits of technology in this field. Therefore, despite the large theoretical basis, neural networks remain a poorly-studied and actual problem. The study of existing means of constructing and training a neural network on the example of a certain strategic game will determine the optimal parameters of its functioning for further use in the gaming industry. The game application to be developed should not be inferior to the human player in making decisions and strategy games, modelling the behaviour and thinking of the person, satisfying the requirements of modern users to the level of artificial intelligence and the likelihood of the game. |
| URI: | http://ea.donntu.edu.ua/jspui/handle/123456789/30369 |
| ISSN: | 1996-1588 |
| Appears in Collections: | Наукові публікації кафедри прикладної математики та інформатики |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| HlynskaKostyukova.pdf | 2,2 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.