A Numerical Method for Solving Fractional Differential Equations by Using Neural Network

Abstract EN

We present a new method for solving the fractional differential equationsof initial value problems by using neural networks which are constructed fromcosine basis functions with adjustable parameters.

By training the neural networksrepeatedly the numerical solutions for the fractional differential equations wereobtained.

Moreover, the technique is still applicable for the coupled differential equationsof fractional order.

The computer graphics and numerical solutions showthat the proposed method is very effective.