Comparative study of integer-order and fractional-order artificial neural networks: Application for mathematical function generation

• DOI: 10.1016/j.prime.2024.100601
• Publication Date: 2024-05-20T16:42:57Z
• Authors: Manisha Premkumar...; Savita Bhosale; Vishwesh A. Vyawa...
• Keywords: Caputo fractional derivative; perceptron; Atangana-Baleanu fractional derivative; 0202 electrical engineering, electronic engineering, information engineering; Riemann-Liouville fractional derivative; Electrical engineering. Electronics. Nuclear engineering; 02 engineering and technology; Grunwald-Letnikov fractional derivative; Caputo-Fabrizio fractional derivative; TK1-9971

This paper investigates the impact of fractional derivatives on activation functions an artificial neural network (ANN). Based results and analysis, a three-layer backpropagation model with integer employed in function gradient descent method learning algorithm has been proposed. Specifically, three perceptrons have proposed based applied to algorithms. They are derivative (FDAF) perceptron, (IDAF)-fractional (FDLA) perceptron. The Riemann-Liouville (RL) derivative, Grunwald-Letnikov (GL) Caputo-Fabrizio (CF) Caputo (C) Atangana-Baleanu (AB) employed. these derivatives' fractional-order (FO) is investigated wide range from 0.1−0.9 testing mean square error (MSE) time required train FO-based IO-based compared help performance metrics such as MSE models. training simulation illustrate that derivative-based outperform other derivatives. Also, perceptron better terms least MSE.