Open AccessOn the Problem of Choosing Optimal Methods for Approximating Functions
Author(s) -
Ilya Bordanov,
S N Zhiganov,
S.N. Danilin
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2096/1/012054
Subject(s) - calculator , computer science , approximation error , function approximation , artificial neural network , decomposition , field (mathematics) , mathematical optimization , function (biology) , feed forward , feedforward neural network , square root , algorithm , mean squared error , mathematics , artificial intelligence , geometry , control engineering , evolutionary biology , pure mathematics , engineering , biology , operating system , ecology , statistics
The materials of the article relate to the field of optimization of control systems and signal processing when preparing models for technical implementation. The informational level of structural and functional decomposition of models of approximators of square root functions is considered. The article investigates two classes of computational methods: sequential - polynomials of the best approximation and parallel - multilayer feedforward neural networks. For each of the classes, using particular examples, the approximation error was calculated according to the criteria of the maximum absolute error and the area of the error function, as well as the computational costs as the sum of the number of mathematical operations and queries in the memory of the calculator.