620 / 2022-03-30 20:02:19

NAS-based Machine Learning Method for Solving Electromagnetic Equations

Electromagnetic Equation,Machine Learning,Partial Differential Equation

Solving electromagnetic equations, a typical class of partial differential equations, is an inevitable problem in the practice of high voltage engineering, such as in multiphysics modelling. Traditional methods for numerically solving PDEs generally encounter bottlenecks in meshing and high-order discretization. Recently, the approaches of using neural networks to solve PDEs have drawn great attention, but the reliance on prior knowledge when designing neural network architectures remains a problem. Therefore, a machine learning method for solving electromagnetic equations based on neural architecture search (NAS) is proposed in this paper.