Computational characteristics of feedforward neural networks for solving a stiff differential equation
نویسندگان
چکیده
Abstract Feedforward neural networks offer a possible approach for solving differential equations. However, the reliability and accuracy of approximation still represent delicate issues that are not fully resolved in current literature. Computational approaches general highly dependent on variety computational parameters as well choice optimisation methods, point has to be seen together with structure cost function. The intention this paper is make step towards resolving these open issues. To end, we study here solution simple but fundamental stiff ordinary equation modelling damped system. We consider two equations by forms. These classic actual method trial solutions defining function, recent direct construction function related method. Let us note settings can easily applied more generally, including partial By very detailed study, show it identify preferable choices made methods. also illuminate some interesting effects observable network simulations. Overall extend literature field showing what done order obtain useful accurate results approach. doing illustrate importance careful setup.
منابع مشابه
A STRONG COMPUTATIONAL METHOD FOR SOLVING OF SYSTEM OF INFINITE BOUNDARY INTEGRO-DIFFERENTIAL EQUATIONS
The introduced method in this study consists of reducing a system of infinite boundary integro-differential equations (IBI-DE) into a system of al- gebraic equations, by expanding the unknown functions, as a series in terms of Laguerre polynomials with unknown coefficients. Properties of these polynomials and operational matrix of integration are rst presented. Finally, two examples illustra...
متن کاملDifferential transformation method for solving a neutral functional-differential equation with proportional delays
In this article differential transformation method (DTMs) has been used to solve neutral functional-differential equations with proportional delays. The method can simply be applied to many linear and nonlinear problems and is capable of reducing the size of computational work while still providing the series solution with fast convergence rate. Exact solutions can also be obtained from the kno...
متن کاملAn exponential spline for solving the fractional riccati differential equation
In this Article, proposes an approximation for the solution of the Riccati equation based on the use of exponential spline functions. Then the exponential spline equations are obtained and the differential equation of the fractional Riccati is discretized. The effect of performing this mathematical operation is obtained from an algebraic system of equations. To illustrate the benefits of the me...
متن کاملNumerical approach for solving a class of nonlinear fractional differential equation
It is commonly accepted that fractional differential equations play an important role in the explanation of many physical phenomena. For this reason we need a reliable and efficient technique for the solution of fractional differential equations. This paper deals with the numerical solution of a class of fractional differential equation. The fractional derivatives are described...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Neural Computing and Applications
سال: 2022
ISSN: ['0941-0643', '1433-3058']
DOI: https://doi.org/10.1007/s00521-022-06901-6