Wolfe既约梯度法及其在权优化中的应用

THE OPTIMUM DESIGN OF OBSERVATION WEIGHT AND ITS STUDY

关于多个约束条件下的观测权的最优化设计,目前大家致力于下面两个方面的研究:一是准则矩阵;二是寻求各种直接解算按纯精度标准建立的数学模型的方法。现在这两个方面尽管都已有不少的方法。但大多数都较处谈且都要做许多准备工作。笔者在研究最优化方法时发现,Wolfe即约梯度法能较好地解算多个约束条件下的观测权的最优化设计,具有准备工作少等优点。本文首先介绍Wolfe既约梯度法,然后较祥细地阐述它在观测权优化设计中的应用。

Now many researchers are studying on the followimg two directions on the optimum de- Slgn of observation weight under multi-constraints:one is normalized matrix,another is to look for the dnect solving processes of mathematics model built in the standard of pure decision Recently,though many methods have been found in those two directions,most of them are haevy and complicated and need many preparations,In the research on optimum programming,the auther discovered that Wolfe's rreducible gradient method can solve the optimum programming of observation Weight under multi- constraints well,and this method has advantages of less preparations and so on.This paper introduces Woife's irreducible gradient method first,then describes how it be applied to the optimum program- ming of observation weight under multi-constraints.