摘要
对水平线性互补问题提出了一种广义中心路径跟踪算法.任意的原始-对偶可行内点均可作为算法的初始点.每步迭代选择"仿射步"与"中心步"的凸组合为新的迭代方向,采用使对偶间隙尽可能减小的最大步长.算法的迭代复杂性为O(nL).
In this paper, a general central path following algorithm for monotone horizontal linear complementarity problem is presented. Any primal-dual feasible interior point can be taken as initial point of the algorithm. At each iteration, we choose a new direction with the convex combination of the affine-scaling step and centering step, and use a pure Newton step with the largest possible reduction in complementarity measure (duality gap). The algorithm maintains the O(√nL ) iteration comolexitv.
出处
《应用数学》
CSCD
北大核心
2011年第2期304-311,共8页
Mathematica Applicata
基金
Supported by the Natural Science Foundation of Hubei Province (2008CDZD47)
关键词
水平线性互补问题
内点方法
广义中心路径跟踪算法
多项式复杂性
Horizontal linear complementarity problem
Interior point method
Gen-eral central path following algorithm
Polynomial complexity