Beyond the first three generations of metamaterials, space-time metamaterials provide the important capability for dynamically manipulating the electromagnetic wave. By controlling the structure parameters, the metamaterial could exhibit the nonrecipr...
Beyond the first three generations of metamaterials, space-time metamaterials provide the important capability for dynamically manipulating the electromagnetic wave. By controlling the structure parameters, the metamaterial could exhibit the nonreciprocal characteristic implying that wave propagates in different ways when interchanging source and observation points. Due to depending on multiple parameters, this characteristic is difficult to investigate. In this paper, we propose a time-varying metamaterial-based waveguide that breaks the reciprocity nature. After that, we apply Deep Neural Network (DNN) algorithm to determine which parameter sets show this property. The results show that DNN successfully predicts the forward and backward transmission difference with high accuracy: 99.2% and 98.8% of 1,472 cases for each have mean square error (MSE) of less than 2.5×10<sup>-3</sup>. In addition, comparing to ADS simulation, the computation time can be decreased 1,500 times.