Title: Stable-plastic neural network, which defines several solutions and new information on its inputs

Authors: Valery Dmitrievich Dmitrienko; Sergey Yurievich Leonov; Alexander Yurievich Zakovorotniy

Addresses: National Technical University, 'Kharkov Polytechnic Institute', 61002, Kharkov, St. Kirpichev, 2, Ukraine ' National Technical University, 'Kharkov Polytechnic Institute', 61002, Kharkov, St. Kirpichev, 2, Ukraine ' National Technical University, 'Kharkov Polytechnic Institute', 61002, Kharkov, St. Kirpichev, 2, Ukraine

Abstract: The analysis results of the discrete Hamming neural network's functioning features are presented. The above-mentioned network cannot retrain in the process of functioning, determine new information on its input layer and recognise input black and white images located at the same minimum distance from two or more reference images. The shortcomings of neural networks using the Hamming distance and solving this problem for images located on the boundaries of two or three classes of images are discussed. A modification of the Hamming neural network which does not have the aforementioned disadvantages was proposed.

Keywords: neural network of Hamming; reference images; images on the boundaries of two or three classes; new information.

DOI: 10.1504/IJCAET.2022.123991

International Journal of Computer Aided Engineering and Technology, 2022 Vol.16 No.4, pp.468 - 477

Received: 16 Apr 2019
Accepted: 27 Oct 2019
Published online: 06 Jul 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article