摘要
在动态测试数据处理中 ,常常要进行稳健回归分析和最小最大值回归分析 .在 Gauss-Newton法的基础上推导了一种用于 lp 数据拟合的算法 .通过对 lp 准则的改造 ,建立了 lp 准则和二次准则之间的关系 .计算结果表明 ,该算法结构清晰 ,便于计算机编程 .当 p=1时 ,可应用于稳健回归分析 ;而当 p>2 0时 ,lp 数据拟合与 l∞ 数据拟合的计算结果十分接近 。
Robust regression analysis and minmax residual error analysis are the two aspects in data processing of dynamic measurement. A new algorithm for l p data fitting based on Gauss-Newton method was deduced. Through establishing the connection between l p criterion and least square criterion, the algorithm has the same property as Gauss-Newton method. It can be applied to the robust regression analysis when p=1. The results of l p data fitting converge to that of l ∞ data fitting when p is a large number (generally p>20) and can be used for the minmax residual error analysis. The experimental results show that the algorithm can be easily realized.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2002年第7期962-965,共4页
Journal of Shanghai Jiaotong University