Approximation of Two-dimensional Numerical Green’s Function Using Neural Network

In this work, a fully-connected forward neural network is constructed to approximate the numerical Green’s function in two-dimensional space where an inhomogeneous region is contained. The outputs predicted by the well-trained neural network and the data calculated by MoM are in good agreement, which approves the effectiveness of the neural network we proposed. And the CPU running time and memory usage of obtaining electric field intensity in testing datasets is much less by using the neural network than MoM.