On the neural network solution of stiff initial value problems

Computational Intelligence techniques are becoming more and more popular in the treatment of classical mathematical problems. The use of Neural Networks (NN) for the solution of Differential Equations is not a new perspective, as this scientific area is active for more than twenty five years. Nevert...

Full description

Saved in:
Bibliographic Details
Main Authors: Kaloutsa, Vaso, Famelis, Ioannis Th
Format: Conference Proceeding
Language:English
Subjects:
Artificial intelligence
Boundary value problems
Differential equations
Mathematical problems
Neural networks
Solvers
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Computational Intelligence techniques are becoming more and more popular in the treatment of classical mathematical problems. The use of Neural Networks (NN) for the solution of Differential Equations is not a new perspective, as this scientific area is active for more than twenty five years. Nevertheless, there are many aspects that can be studied in both theoretical and technical point of view. In this work we investigate the NN based approximate solution of stiff Initial Value Problems (IVPs). We extend the work of Lagaris et al [1] and propose an alternative NN architecture and activation function choices. In order to reveal the capabilities of NN solutions we compare our solutions to the solutions of standard MATLAB stiff solvers.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0026823