【摘要】 本文提出一种解线性约束凸规划的数值方法。通过将问题的KKT系统转化成一个约束方程,算法在每步迭代只需解一个线性方程组即可得到搜索方向。算法运用了信赖域方法利内点技术。在较弱的条件下,我们证明了算法的全局收敛性。更多还原

【Abstract】 This paper presents an interior trust region method for linear constrained LC convex optimization problems.By reformulating the KKT system as a constrained equation,the algorithm generates the search direction by solving a linear equation at each iteration.The algorithm combines trust region method and interior technique.Under mild conditions,we show that the algorithm is globally convergent.更多还原