解非线性最小二乘问题的常用算法是高斯-牛顿法。它的缺点是迭代的每一步都要进行逆矩阵运算,对参数初估值的要求也比较严。I.A.Slavic提出的不用逆矩阵运算的方法(简记为AWMI法),已被金慧娟等用于穆斯堡尔谱的拟合,认为收敛性很好,计算量也较小。 本文通过理论分析说明按AWMI法的参数修正方向比较接近于x~2量的最快速下降方向,因此在步长不太 A common algorithm for solving nonlinear least squares problems is the Gauss-Newton method. Its disadvantage is that each iteration of the iteration should be carried out inverse matrix operation, the initial evaluation of the parameters of the more stringent requirements. I.A.Slavic proposed method without inverse matrix operation (abbreviated as AWMI method), has been used by Huijin Juan Mossbauer fitting, that the convergence is good, the calculation is also small. In this paper, the theoretical analysis shows that the direction of parameter correction according to the AWMI method is closer to the fastest descent direction of x ~ 2, so the step size is not too