Алгоритмы нейросетевого моделирования многомерных образов на базе неархимедовых алгебраических исчислений

Abstract

Recurrent synthesis algorithms of multy-valued polynomial and threshold polynomial neural network for modeling multivariate visual patterns are constructed. Recurrent synthesis algorithms of multy-valued neural networks, which extends neural networks of polynomial and threshold polynomial type for modeling multivariate visual patterns are constructed. Constructed algorithms allows to improve the efficiency of algebraic approach to pattern recognition and image analysis problems.