Robust估计的算法优化

本文分析了Ｒｏｂｕｓｔ估计收敛慢的原因，提出以最小二乘平差的残差（Ｖ）为初始信息，用Ｌ１估计的权函数所计算的权力初始权，根据迭代计算的残差的大小，及残差的变化所提供的信息，构造出加速收敛的算法，计算结果表明，较之一般的方法，迭代次数显著减少，辨识杠杆点含粗差的能力有较大提高，即搜索粗差的效率有很大提高。

THE OPTIMIZATION OF ROBUST ESTIMATION COMPUTING METHOD

This paper analyses the causes in convergence slowly of Robust estimation,advances residual (v)which is gained from the least squares adjustment computation as the initial information,uses the weight function of L 1 estimation to compute and obtain the weight as the initial weight,according to residual size and residual change,computing method is consturcted to accelerate convergence.Computation results indicate that the alternative computation times are notably reduced,the capacity of identifying leverage point which involves gross errors is relatively raised,it means that the efficiency of searching gross errors has improved a lot.