Facts and issues of neural networks for numerical simulation - Université Paris Lumières
Book Sections Year : 2024

Facts and issues of neural networks for numerical simulation

Abstract

Deep learning and artificial intelligence (AI) have transformed computer science, becoming the main method for addressing various problems, from partial differential equations to molecular discovery. Traditional numerical simulation techniques use finite differences to solve equations, but are computationally costly, often taking hours to days on powerful machines. Simulations need to be restarted for shape changes, making the process inefficient. Artificial neural networks now enable accurate simulations quickly and inexpensively. This paper reviews recent developments in neural networks, their advantages, disadvantages, and unresolved issues, often demonstrated through partial differential equations.
Fichier principal
Vignette du fichier
Neural_networks_for_numerical_simulation.pdf (479.39 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04857297 , version 1 (28-12-2024)

Licence

Identifiers

  • HAL Id : hal-04857297 , version 1

Cite

Imad Kissami, Christophe Cérin, Fayssal Benkhaldoun, Fahd Kalloubi. Facts and issues of neural networks for numerical simulation. Mostapha Zbakh, Mohammed Essaaidi, Claude Tadonki, Abdellah Touha and Dhabaleswar K. Panda. Artificial Intelligence and High-Performance Computing in the Cloud - Research and Application Challenges., Lecture Notes in Networks and Systems, In press. ⟨hal-04857297⟩
0 View
0 Download

Share

More