Deep Conditional Generative Adversarial Network-Assisted Inverse Design of Metasurface

Conventional metasurface (MS) design relies on time-consuming simulation iterative method and high design costs. As a new technology, deep learning (DL) has been applied in the field of MS design and shown good performance. This paper develops an innovative deep conditional generative adversarial network (DCGAN) for the inverse design of MS. By constructing and training the network model, the corresponding MS structure can be quickly and automatically generated with the required polarization conversion rate (PCR) as input, and then the purpose of on-demand design can be achieved.