A Data Augmentation Method for Antenna Optimization

To alleviate the problem of training data scarcity, a data augmentation method consisting of a conditional generative adversarial network (CGAN) and an encoder is proposed in this study. The encoder can provide the prior knowledge with the CGAN, and then the CGAN can generate geometric parameters and corresponding responses. Based on the generated data, an inverse random forest (RF) and a forward RF are built to obtain the optimized S-parameters. A broadband stacked patch array antenna is used to verify our method, and the results demonstrate that the established surrogate models are accurate, and an isolation of −43.70 dB can be achieved.