Inverse design of multiparameter antenna using hybrid machine learning‐driven training dataset

Abstract

Conventional approaches for designing antennas are often time‐consuming and computationally expensive processes. In this letter, a time and resource‐efficient inverse design method for the multiparameter antenna is proposed which is capable of generating an effective dataset using the hybrid machine learning method. The proposed method filters the design variables and predicts the secondary variables based on the choice of primary design variables. The generated training dataset is used for the effective prediction of design variables based on multiple performance metrics. The proposed machine learning model utilized autoencoder for dimensionality reduction of multiple performance metrics and support vector regression is used for the prediction of design variables. To prove the effectiveness of the proposed model, a metasurface‐loaded low‐profile antenna is considered as a design example.