Abstract
Hopfiled neural network (HNN) is considered as a means for processing data obtained by ionosphere measurements. Data look like tracks of a complicated structure with a considerable background. That required preliminary data filtering and compression. Both are accomplished by applying a cellular automation and median smoothing. The specially modified rotor HNN model is used. For constructing its initial configuration an angle histogramming is applied in a domain which size is determined by the mean local track curvature. Neuron weight functions are constructed as a measure of the direction closeness of rotors and their tracks. The proper choice of the initial configuration guarantees the network fast convergence to the stable state, corresponding to the global minimum of its energy function. Mean errors of the track resulting identification are about 5%.(author). 12 refs.; 6 figs.
Zaznobina, E G;
[1]
Ososkov, G A
[2]
- Ivanovskij Gosudarstvennyj Univ., Ivanovo (Russian Federation)
- Joint Inst. for Nuclear Research, Dubna (USSR). Lab. of Computing Techniques and Automation
Citation Formats
Zaznobina, E G, and Ososkov, G A.
Neural network application for data analysis of vertical ionosphere locating; Primenenie nejronnykh setej v analize dannykh vertikal`nogo zondirovaniya ionosfery.
JINR: N. p.,
1993.
Web.
Zaznobina, E G, & Ososkov, G A.
Neural network application for data analysis of vertical ionosphere locating; Primenenie nejronnykh setej v analize dannykh vertikal`nogo zondirovaniya ionosfery.
JINR.
Zaznobina, E G, and Ososkov, G A.
1993.
"Neural network application for data analysis of vertical ionosphere locating; Primenenie nejronnykh setej v analize dannykh vertikal`nogo zondirovaniya ionosfery."
JINR.
@misc{etde_10144864,
title = {Neural network application for data analysis of vertical ionosphere locating; Primenenie nejronnykh setej v analize dannykh vertikal`nogo zondirovaniya ionosfery}
author = {Zaznobina, E G, and Ososkov, G A}
abstractNote = {Hopfiled neural network (HNN) is considered as a means for processing data obtained by ionosphere measurements. Data look like tracks of a complicated structure with a considerable background. That required preliminary data filtering and compression. Both are accomplished by applying a cellular automation and median smoothing. The specially modified rotor HNN model is used. For constructing its initial configuration an angle histogramming is applied in a domain which size is determined by the mean local track curvature. Neuron weight functions are constructed as a measure of the direction closeness of rotors and their tracks. The proper choice of the initial configuration guarantees the network fast convergence to the stable state, corresponding to the global minimum of its energy function. Mean errors of the track resulting identification are about 5%.(author). 12 refs.; 6 figs.}
place = {JINR}
year = {1993}
month = {Dec}
}
title = {Neural network application for data analysis of vertical ionosphere locating; Primenenie nejronnykh setej v analize dannykh vertikal`nogo zondirovaniya ionosfery}
author = {Zaznobina, E G, and Ososkov, G A}
abstractNote = {Hopfiled neural network (HNN) is considered as a means for processing data obtained by ionosphere measurements. Data look like tracks of a complicated structure with a considerable background. That required preliminary data filtering and compression. Both are accomplished by applying a cellular automation and median smoothing. The specially modified rotor HNN model is used. For constructing its initial configuration an angle histogramming is applied in a domain which size is determined by the mean local track curvature. Neuron weight functions are constructed as a measure of the direction closeness of rotors and their tracks. The proper choice of the initial configuration guarantees the network fast convergence to the stable state, corresponding to the global minimum of its energy function. Mean errors of the track resulting identification are about 5%.(author). 12 refs.; 6 figs.}
place = {JINR}
year = {1993}
month = {Dec}
}