Available online

  • Articles 3759
  • Volumes 11
  • Issues 120

UDK: 681.3.06:004.8

2017. - V. 31. - N 8(189). - P. 57-59

I.S. Markin, M.D. Pysin, D.O. Krasnov, S.P. Dudarov,

ALGORITHMIC SOFTWARE FOR APPROXIMATION OF NONLINEAR RELATIONS BY NEURAL NETWORK WITH RADIAL-BASIS FUNCTIONS

Algorithmic software for approximation of nonlinear dependencies, that simulates the operation of artificial neural network based on radial-basis functions, was developed. Functional capabilities of the software for obtaining an approximating mathematical model and visualization of results, obtained in form of tables and graphs, was presented. An example of approximation of a one-dimensional composite function. Impact of parameter of activation radial-basis function and count of hidden radial elements in structure of neural network on value of approximation error was studied. A significant error decrease was shown with increase in count of radial elements and optimal choice of saturation parameter for Gaussian function.

Get PDF

First page of the article:

ПРОГРАММНО-АЛГОРИТМИЧЕСКОЕ ОБЕСПЕЧЕНИЕ ДЛЯ АППРОКСИМАЦИИ НЕЛИНЕЙНЫХ ЗАВИСИМОСТЕЙ НЕЙРОННОЙ СЕТЬЮ НА ОСНОВЕ РАДИАЛЬНО-БАЗИСНЫХ ФУНКЦИЙ

References

  1. 1

Keywords

algorithmic software   approximation error   approximation of nonlinear relations   artificial neural network   Gaussian function   radial basis function   radial element   RBF-network   аппроксимация нелинейных зависимостей   искусственная нейронная сеть   ошибка аппроксимации   программно-алгоритмическое обеспечение   радиально-базисная функция   радиальный элемент   РБФ-сеть   функция Гаусса  


Journal indexing

  • Elibrary.ru
  • Винити

Our friends

  • Фонд "Научная перспектива"
  • Минобрнауки
  • РХО им. Д.И. Менделеева