贝叶斯优化的学生T过程用于微带天线谐振频率建模

Resonant frequency modelling of microstrip antenna based onBayesian optimization students′T process

学生T过程的核心是核函数设计以及超参数优化,其中超参数直接影响模型的泛化能力.为提升学生T过程模型的预测精度,基于一种自适应的贝叶斯优化算法,实现对学生T过程超参数的全局优化.以多组Benchmark问题和矩形微带天线谐振频率建模为仿真实例,并与多种机器学习模型进行了对比实验.结果表明:采用贝叶斯优化超参数的学生T过程具有较高的拟合精度.

The key issues of students′T process(STP)are kernel function design and hyperparameter optimization,and the optimized hyperparameter affects directly the generalization ability of the model.In order to improve the prediction accuracy of STP,this study introduces an adaptive Bayesian optimization(BO)algorithm to optimize the hyperparameters of STP.Taking Benchmark questions and resonant frequency modeling of microstrip antenna as examples,the experimental comparisons are conducted with several machine learning models.This research results show that STP optimized by BO has higher fitting accuracy.