求解最小<i>L</i><sub>1</sub>-模估计的数学规划方法

求解最小L₁-模估计的数学规划方法

摘要: 最小L₁-模估计(也称为最小—乘估计)在回归分析中有着十分重要的意义,但其计算确相当困难.以致于影响到它的应用.本文就最小L₁-模估计的计算,介绍几种求解的数学规划方法,它们包括:(1)单纯形方法;(2)L₁-模估计的对偶规划方法;(3)投影梯度方法;(4)有效集法.

Abstract: Minimum L₁-norm Estimation is very important on theory of linear regression analysis. It is quite difficult to solve. This paper surveys some methods for solving minimum L₁-norm Estimation through mathematical programming, which include (1)simplex method, (2) dual linear programming, (3) gradient projection method and (4) active set method.