Optimization of Neural Network Identification of a Non-Stationary Object Based On Spline Functions

Abstract

A technique for smoothing a dynamic process based on basis-spline functions and calculating information recovery coefficients has been developed, which helps to optimize the training of a neural network data processing system by reducing the errors of the training subset. Methods and algorithms for modeling the processes of smoothing, processing, and restoring data of non-stationary processes based on cubic spline functions are studied.