Original A new modern scheme for solving fractal-fractional differential equations based on deep feedforward neural network with multiple hidden layer

No Thumbnail Available

Date

2024

Journal Title

Journal ISSN

Volume Title

Publisher

Elsevier

Open Access Color

OpenAIRE Downloads

OpenAIRE Views

Research Projects

Organizational Units

Journal Issue

Events

Abstract

The recent development of knowledge in fractional calculus introduced an advanced superior operator known as fractal-fractional derivative (FFD). This operator combines memory effect and self-similar property that give better accurate representation of real world problems through fractal-fractional differential equations (FFDEs). However, the existence of fresh and modern numerical technique on solving FFDEs is still scarce. Originally invented for machine learning technique, artificial neural network (ANN) is cutting-edge scheme that have shown promising result in solving the fractional differential equations (FDEs). Thus, this research aims to extend the application of ANN to solve FFDE with power law kernel in Caputo sense (FFDEPC) by develop a vectorized algorithm based on deep feedforward neural network that consists of multiple hidden layer (DFNN-2H) with Adam optimization. During the initial stage of the method development, the basic framework on solving FFDEs is designed. To minimize the burden of computational time, the vectorized algorithm is constructed at the next stage for method to be performed efficiently. Several example have been tested to demonstrate the applicability and efficiency of the method. Comparison on exact solutions and some previous published method indicate that the proposed scheme have give good accuracy and low computational time.

Description

Ahmadian, Ali/0000-0002-0106-7050; Salahshour, Soheil/0000-0003-1390-3551

Keywords

Fractal-fractional differential equation, Artificial Neural Network, Deep feedforward neural network, Vectorized algorithm, Adam optimization

Turkish CoHE Thesis Center URL

Fields of Science

Citation

WoS Q

Q1

Scopus Q

Q1

Source

Volume

218

Issue

Start Page

311

End Page

333