Please use this identifier to cite or link to this item:
https://dspace.ncfu.ru/handle/20.500.12258/19609
Title: | Hamming neural network in discrete form |
Authors: | Shaposhnikov, A. V. Шапошников, А. В. Orazaev, A. R. Оразаев, А. Р. Eremenko, E. N. Еременко, Е. Н. Malakhov, D. V. Малахов, Д. В. |
Keywords: | Discretization;Hamming neural network;Integer form;Neural networks;Pattern recognition |
Issue Date: | 2022 |
Publisher: | Springer Science and Business Media Deutschland GmbH |
Citation: | Shaposhnikov, A., Orazaev, A., Eremenko, E., Malakhov, D. Hamming neural network in discrete form // Lecture Notes in Networks and Systems. - 2022. - Том 424. - Стр.: 11 - 17. - DOI10.1007/978-3-030-97020-8_2 |
Series/Report no.: | Lecture Notes in Networks and Systems |
Abstract: | Hamming artificial neural network is used to solve problems of classification of binary input vectors. Its work is based on procedures aimed at choosing, as a solution to the classification problem, one of the reference images closest to the noisy input image supplied to the network input and assigning this image to the corresponding class. At the same time, difficulties were identified in implementing calculations of the Hamming neural network paradigm. One of the ways to simplify computations in the considered neural network paradigm is to transition to integer computations since integer multiplication is computed several times faster than real data type. To simplify the computation model, a Hamming neural network in discrete form was proposed in this paper. A discrete model of the Hamming neural network is proposed, making it possible to simplify the implementation of computations significantly. These results can be applied to solve various problems, including computer vision applications. |
URI: | http://hdl.handle.net/20.500.12258/19609 |
Appears in Collections: | Статьи, проиндексированные в SCOPUS, WOS |
Files in This Item:
File | Size | Format | |
---|---|---|---|
scopusresults 2177 .pdf Restricted Access | 63.23 kB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.