Abstract
Ordinary Differential Equations (ODEs) play a crucial role in various scientific and professional domains for modeling dynamic systems and their behaviors. While traditional numerical methods are widely used for approximating ODE solutions, they often face challenges with complex or nonlinear systems, leading to high computational costs. This study aims to address these challenges by proposing an artificial neural network (ANN)-based approach for solving first-order ODEs. Through the introduction of the ANN technique and exploration of its practical applications, we conduct numerical experiments on diverse first-order ODEs to evaluate the convergence rate and computational efficiency of the ANN. Our results from comprehensive numerical tests demonstrate the efficacy of the ANN-generated responses, confirming its reliability and potential for various applications in solving first-order ODEs with improved efficiency and accuracy.
Reference14 articles.
1. Mahata G., Raut D.S., Parida C., Baral S., & Mandangi S. (2022). Application of First-Order Differential Equations. International Journal of Engineering Science Technologies, 6(5), pp. 23-33.
2. Lu L., Dagger, X. M., Dagger, Z. M., George, S. E. K. (2021). A Deep Learning Library for Solving Differential Equations. Society for Industrial and Applied Mathematics, 63 (1), pp. 208–228.
3. Mao Z., Jagtap, A. D. & Karniadakis, G. E. Physics-informed neural networks for highspeed flows. Computational Methods in Applied Mechanics and Engineering, 360, pp. 209-225.
4. Ji, X.A., Moln´ar, T.G., Avedisov, S.S., & Orosz, G. (2020). Feed-forward neural networks with trainable delay. Learning for Dynamics and Control, 120(1), pp. 127–136.
5. Turan, E.M., & J¨aschke, J. (2021). Multiple shooting for training neural differential equations on time series. Control Systems Letters, 2(6), pp. 1897–1902.