Modeling EM Problem with Deep Neural Networks

Abstract

This paper investigates the potential of using deep neural network (DNN) to model electromagnetic forward problems. As a preliminary attempt, we use a deep convolutional neural network (CNN) to fit the scattered field of an inhomogeneous circular region as calculated by a 2D Finite Element-Boundary Integral (FE-BI) model. This approach provides a new tool to fast map input to output of a specific EM problem, which builds basis for further study on solving inverse problem with DNN.