Кваліфікаційні роботи здобувачів вищої освіти кафедри авіоніки та систем управління

Permanent URI for this collectionhttp://er.nau.edu.ua/handle/NAU/58725

Browse

Browsing Кваліфікаційні роботи здобувачів вищої освіти кафедри авіоніки та систем управління by Subject "AdaBoost algorithm"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Ensemble classifier based on boosting
    (National Aviation University, 2022-11-23) Plodistyi, Bogdan; Плодистий, Богдан Олександрович
    This paper considers the construction of a classifier based on neural networks, nowadays AI is a major global trend, as an element of AI, as a rule, an artificial neural network is used. One of the main tasks that solves the neural network is the problem of classification. For a neural network to become a tool, it must be trained. To train a neural network you must use a training sample. Since the marked training sample is expensive, the work uses semi-supervised learning, to solve the problem we use ensemble approach based on boosting. Speaking of unlabeled data, we can move on to the topic of semi-supervised learning. This is due to the need to process hard-to-access, limited data. Despite many problems, the first algorithms with similar structures have proven successful on a number of basic tasks in applications, conducting functional testing experiments in AI testing. There are enough variations to choose marking, where training takes place on a different set of information, the possible validation eliminates the need for robust method comparison. Typical areas where this occurs are speech processing (due to slow transcription), text categorization. Choosing labeled and unlabeled data to improve computational power leads to the conclusion that semi-supervised learning can be better than teacher-assisted learning. Also, it can be on an equal efficiency factor as supervised learning. Neural networks represent global trends in the fields of language search, machine vision with great cost and efficiency. The use of "Hyper automation" allows the necessary tasks to be processed to introduce speedy and simplified task execution. Big data involves the introduction of multi-threading, something that large companies in the artificial intelligence industry are doing.