Development of prognostic neural network models based on non-positional coding for automatic control systems

Abstract

The article is devoted to the discussion of the problems of development of predictive neural network models based on the residual number system and the use of modular arithmetic to improve the quality of automatic control systems by adding to the control algorithms a prognostic component, which is especially important for astatic control objects. The possibility of implementing neural network training algorithms in the residual number system is shown, which allows to significantly accelerate the work of these algorithms, which is especially important when adding new functionality to automatic control systems in the form of prognostic neural network models