Arutyunyan A. The system of soft calculations on the basis of adaptive resonance neural networks for solving CAD problems.

Thesis for taking the Candidate of Technical Science degree in 05.13.12 - Design works automation systems. - National Technical University of Ukraine "Kyiv Polytechnic Institute", ME of Ukraine, Kyiv, 2009 Thesis work is dedicated to creating hybrid system of soft calculations on the basis of adaptive resonance theory neural network for improving the quality of solving CAD problems in conditions of accuracy, quickness, noise immunity and computing requirements. The architecture and functional model of new adaptive resonance hybrid neural network are developed and theoretically justified in the thesis. Learning technologies are developed and appropriately theoretically justified. The new type of Tychonoff functional using Lipschitzian structural and signal constants is proposed. It allows to perform theoretical research of soft calculations systems' attributes. It's suggested to use hybrid method of nonlinear data normalization for normalization of expert systems' and neural networks' input data. The method is dual.