Ilipbaeva L.B., Utebaeva D.Z.

 

ANALYSIS OF MATHEMATICAL MODELS OF NEURAL NETWORKS IN ENCODING INFORMATION

 

KEY WORDS: NEURAL NETWORK, MATHEMATICAL MODELS, CODING .

 

Currently, neural networks represent a future technology that enables to solve various problems in diverse fields of human activity.

This technology of calculations, modeling analysis provides new approaches to the study of difficult tasks in the telecommunications industry, medicine, economics and it is the actual direction of modern science.

Application of neural networks in telecommunication systems are due to their following advantages:

- many parallel processing of information;

- the ability to get the result in complex, intractable problems.

Neural networks make it possible to model and analyze nonlinear processes. Their ability to work with large amounts of data and make it possible to apply adaptable neural network for solving various problems in telecommunication systems. In recent years, based on neural networks have been proposed and developed many models for use in routing topics, distribution channels in mobile radio systems and recognition of speech signals and transmitting information.

The aim of the this work is the use of the classical models of neural network in coding theory, and to study the neural network model's work. The most pressing problem in the transmission of information in telecommunication systems is the accuracy of the transmitted and received data. For this purpose is used a quantity of methods of error-correcting codes: cyclic codes, convolution codes, etc.

Commonly used codes in the transmission of information theory are cyclic codes. These codes are popular because of their ease of engineering implementation of decoders, as well as the high noise immunity. From the classical sources [1,2] it is known that cyclic codes can detect five or more errors.

Cyclic codes are class of systematic block codes. In cyclic codes, each combination is coded independently in the form of blocks of images of polynomials [Figure 1]. Information and control characters are always at certain places. The basic formula is the construction of cyclic codes:

 

F(X)=U(X)P(X)=G(X)Xm +R(X)

 

where

 which reflects both the method of constructing cyclic code.

 

 

 

 

 


                                                                 

 

 

Figure 1 - the block scheme of model

 

 

This code is used when the data packet. Upon receipt of the packet, the received packet is divided by the generator polynomial. If the result is "0" then the combination adopted without distortion, in the opposite case, there is the "1", then the package is erroneous.

In this article we discussed about analysis of mathematical models of neural networks in encoding information. Much commonly models of neural networks in encoding information are cyclic codes, convolution codes, etc. The advantage of codes: each combination is coded independently in the form of blocks of images of polynomials.

 

 

THE REFERNCES:

 

 

1.      Комашинский В.И., Смирнов Д.А. Нейронные сети и их применение в системах управления и связи. М.: Горячая линия – Телеком, 2003. 94 С., 2002.

2       Хайкин Саймон  нейронные сети : М.: Санкт-Петербург.:-Киев.: 2006.