Vector quantization algorithms for one-dimensional and two-dimensional time series

Abstract

The purpose of this investigation is determining the parameters of non-stationary models for different kind of non-stationary processes. With the help of neural network and vector quantization is suggested an approach for modelling one-dimensional and two-dimensional time series. The parameters of vector quantization are determined for describing probability density function. Programs on MATLAB 6.5 and Visual C++ were created for modelling of time series and receiving their probability density function. These models can be used for describing and analyses the seismic waves behaviour, for queuing models in communication systems and for different kind of models of time series with long-range dependences.