Authors: Sailendra Tyagi, Chukesh Mishra
Abstract: The increasing complexity of scientific and engineering problems necessitates the development of efficient and accurate numerical solvers. Traditional numerical solvers, while reliable, often face challenges with high-dimensional systems, nonlinearity, and computational cost. Recently, the integration of machine learning (ML) techniques into numerical methods has emerged as a promising approach to enhance solver performance. This review presents a comprehensive analysis of machine learning-enhanced numerical solvers, focusing on their principles, methodologies, applications, and limitations. Techniques such as neural network-based solvers, physics-informed neural networks, and surrogate models are discussed in detail. Additionally, the paper explores comparative studies demonstrating the computational advantages and accuracy improvements achieved by ML-assisted solvers. The review concludes by highlighting current challenges, future research directions, and potential interdisciplinary applications.
Keywords: Machine learning, numerical solvers, physics-informed neural networks, surrogate models, computational efficiency, high-dimensional systems
Full Issue
| View or download the full issue | PDF 39-51 |