INCREASING THE EFFICIENCY OF NEURAL NETWORKS IN RECOGNITION PROBLEMS
DOI:
https://doi.org/10.36773/1818-1112-2022-129-3-5-8Keywords:
neural networks, convolution neural network, neuroevolutionary learning, image recognition, character recognition, neural network learning, deep learning, reinforcement learning, extrapolation learning, positional coding, configuration coding, single coding, multiple coding, input coding, output codingAbstract
The article describes the issues of increasing the efficiency of neural networks in terms of their design and coding of input and output signals. The application of multiple signal coding using extrapolation of the input parameters is described on the example of a system of recognition character sequences on images of arbitrary size with a complex background.
An effective combination of multiple positional and configuration-competitive coding for various types of signals makes it possible to achieve performance rates of the building number recognition algorithm of up to 74 images per second in the adaptive learning mode and 218 images per second in the recognition only mode.
The paper also outlines general recommendations for signal coding in artificial intelligence systems based on neural networks.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2022 Brest State Technical University
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
The work is provided under the terms of Creative Commons public license Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). This license allows an unlimited number of persons to reproduce and share the Licensed Material in all media and formats. Any use of the Licensed Material shall contain an identification of its Creator(s) and must be for non-commercial purposes only. Users may not prevent other individuals from taking any actions allowed by the license.
