An artificial neural network approach for a single parameter boundary value problem with singular perturbation

Abstract

This article explores the artificial neural networks(ANN) approach in solving a one parameter singular perturbation boundary value problem. Solution behaviour of a singular perturbation convection-diffusion boundary value problem through a small perturbation restriction is analysed using neural networks. Feed-forward neural networks(FFNN) computations are utilised to determine and investigate the behaviour of proposed solution. A class of neural network is employed for the test functions and the adams approximation optimization system is used to estimate the adjustable factors. Three computational models are conducted to exhibit the efficacy of the technique, and its maximum magnitude errors are matched with those of other approaches in the works in order to validate the approach. Visualized the result profiles for both artificial neural network and analytical methods. The high R-Squared values were attained for prediction. The results imply that artificial neural networks offer an adaptable and efficient strategy for certain classes of singular perturbation boundary value problems, supporting further study and practical use.