High efficiency design of metal–insulator–metal metasurface by ResNets-10

Abstract

Deep learning prediction of metasurface has been a widely discussed issue in recent years. However, the prediction accuracy is still one of the challenges to be solved. In this work, we proposed using the ResNets-10 model to predict plasmonic metasurface S11 parameters. The two-stage training was performed by the k-fold cross-validation and small learning rate. After the training was complete, the predicted logarithmic losses for aluminum, gold, and silver metal–insulator–metal metasurfaces were −48.45, −46.47, and −35.54, respectively. Due to the ultralow error value, the proposed network can efficiently replace the traditional computing methods within a certain structural range. The ResNets-10 can complete training within 1100 iterations, which is highly efficient. The ResNets-10 model we proposed can also be used to design meta-diffractive devices and meta-resonance biosensors, thereby reducing the time required for the simulation process. The ultralow lose value of the network indicates that this work contributes to the development of future artificial intelligence electromagnetic devices computing software.