一种基于神经网络的高频电阻建模方法  

A modeling method of high frequency resistance based on neural network

作　　者：李志强 任坤 刘军[1] LI Zhiqiang;REN Kun;LIU Jun(Key Laboratory of RF Circuits and Systems,Ministry of Education,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)

机构地区：[1]杭州电子科技大学浙江省大规模集成电路设计重点实验室,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2022年第3期8-12,35,共6页Journal of Hangzhou Dianzi University：Natural Sciences

摘　　要：提出一种基于神经网络的高频电阻全局模型。利用先验信息进行数据扩充,并建立特征工程,解决了小样本学习中过拟合问题,提高了神经网络的泛化能力。与传统建模方法相比,提出的基于神经网络的高频电阻建模方法降低了模型表征的误差,提升了建模效率。A global model of high frequency resistance based on neural network is proposed.Using prior information to expand data and establish feature engineering,the over fitting problem in small sample learning is solved,and the generalization ability of neural network is improved.Compared with the traditional modeling methods,the proposed high-frequency resistance modeling method based on neural network can reduce the error of model representation and improve the modeling efficiency.

关 键 词：神经网络 器件建模 小样本学习 

分 类 号：TN713.5[电子电信—电路与系统]