摘要
放大器的广泛适用性决定了系统实现的准确性。运用支持向量回归机,结合光滑技术的优越性,利用Newton算法使约束化问题有效转化为无约束化简化问题,给出了基于支持向量回归调节熵函数法的设计思想及构造步骤。实验以高校模拟电子技术实验为依托,采用近一年内由精密仪器设备测评所得的小功率放大器的8项指标构建训练集,进行支持向量回归调节熵的评价。实验表明,与传统方法相比,该方法可以有效地避免为追求精确解,而令极大熵函数中参数取值过大而造成的病态矩阵或数值溢出现象,同时实现了收敛速度快、存储空间小的预期,可推广于电子产品的性能评价中去。
The general applicability of amplifier decides the accuracy of the system implementation.Combining the superiority of smoothing technique with support vector regression machine,Newton algorithm is used to translate the constrained problem into unconstrained minimization simple problem effectively.The design idea and construction steps for the adjustable entropy function method based on support vector regression are introduced.Experiments are carried out,which are supported by college analog electronic technology experiments;eight low power amplifier performance specifications obtained from precision instrument evaluation results in one year are used to construct the training set.Then the experiments are used to evaluate support vector regression adjustable entropy function（AEFSVR）.Experiment results show that this method could avoid the phenomenon,such as ill-conditioned matrix and numerical overflow in the maximum entropy function with a relatively small parameter,realizes the anticipation of fast convergence speed and small storage space.This method is better than traditional method and suitable for the comprehensive evaluation of electronic products.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2011年第2期295-301,共7页
Chinese Journal of Scientific Instrument
基金
辽宁省教育厅(2009A045)资助项目
关键词
支持向量回归机
调节熵函数
放大器性能
Newton算法
support vector regression machine
adjustable entropy function
amplifier performance
Newton algorithm
