A Metasurface Modeling Method Based on Generative Adversarial Network Combined with K-Nearest Neighbor

Abstract

A machine learning-based metasurface modeling method using generative adversarial network (GAN) and k-nearest neighbor (k-NN) is proposed. The k-NN algorithm is used for classification of metasurfaces samples, and GAN is used for the design of metasurfaces. By applying the machine learning-based modeling method, GAN models for designing metasurfaces are obtained. Simulation results show that using the GAN models combined with k-NN can design metasurfaces more efficiently.