Yurchak I. Synthesizing and applying the neural networks with projective and ordered lateral links

The architecture of neural networks with projective and ordered lateral links built on the basis of new conceptual model "The Functional on set of tabulared functions" is picked out as a subject of researches in this thesis work. The version of polynomial model which integrates the interpolating and the approximating approaches of tabulared relations reconstruction is elaborated on the basis of researches of miscellaneous versions of construction of transfer functions of neural elements in the hidden layer. The practical training and operating algorithms of these neural networks have been designed here and the characteristics of neural networks have been estimated at implementation of the class of data mapping problems. The practical guidelines of the data pretreatment for practical use have been elaborated here. The obtained results have passed check in conditions of real life on practical tasks of electric power industry and scientific researches