An Efficient Artificial Neural Network Model for Inverse Design of Metasurfaces
Lin Yuan, Lan Wang, Xue‐Song Yang, Hao Huang, Bing‐Zhong Wang
Abstract
To expedite the design process of metasurface, an improved transfer function (TF)-based artificial neural network (ANN) model is proposed, which can directly generate structure parameters to match the customer-expected electromagnetic (EM) response. Compared with the existing inverse design techniques, poles and residues of TF instead of the EM responses at discrete points are input into the proposed model. To identify the solution to circumvent the nonuniqueness problem, which is the major challenge of inverse design, a novel network structure and a new training error function are proposed. An electromagnetically induced transparency-like metasurface is selected as an example to verify the effectiveness and efficiency of the proposed inverse design model.