Problems of Further Development of GMDH Algorithms: Part 2
Theories and algorithms developed for pattern recognition can be applied to random processes forecasting and for solution of all another interpolation type problems of artificial intelligence. For this purpose input data sample in the form of time series should be transformed into simultaneous form according to rules of Gauss conditional equations complication. Examples of algorithms used in pattern recognition are considered. Particularly is considered algorithm of secondary arguments generation. It is proposed to use error of modelling as effective secondary argument in special twice-multilayered neural network. Another GMDH network is considered as inductive analogue of Kalman type noise filter and as network, which interpolates non-linear objects characteristics.
A. G. Ivakhnenko et al., "Problems of Further Development of GMDH Algorithms: Part 2," Pattern Recognition and Image Analysis, MAIK/ Pattern Recognition and Image Analysis, Jan 2002.
Electrical and Computer Engineering
Article - Journal
© 2002 MAIK/ Pattern Recognition and Image Analysis, All rights reserved.
01 Jan 2002