基于两阶段高斯过程的天线建模方法

Antenna modeling method based on two-stage Gauss process

针对高斯过程机器学习解决电磁问题,提出了两阶段高斯过程的天线建模方法,共包含两个阶段,在第一阶段,学习天线的粗细模型之间的映射关系,从而在第二阶段建立起高精度细模型的实际代替模型,在降低天线高精度输入训练数据的计算代价上效果显著。将两阶段高斯过程建模方法应用在倒F天线的优化问题和双频PIFA天线的谐振频率预测问题中,通过选取细模型数据占总训练数据的不同比例,比较它们的多种误差,从而验证该两阶段高斯过程天线建模方法的有效性和准确性。

Gauss process(GP)as a machine learning method solves electromagnetic problems.This paper presented the me-thod of antenna modeling based on two-stage GP.The method consisted of two stages.In the first stage,it studied the mapping between the antenna models of different thicknesses.In the second stage,it established the practical model of high precision fine model.It had significant effects on reducing the computational cost of high precision input training data.The two-stage GP modeling method solved the optimization of the inverted-F antenna and the prediction of resonant frequency of dual frequency PIFA antenna.This paper selected the different proportion of the fine model data to the total training data to compare their different errors,so that it verifies the validity and accuracy of the antenna modeling method based on the two-stage GP.