文章对IPA(inerior point algorithm)进行了修正,证明修正的IPA的收敛性;将修正的IPA应用于不等式约束凸优化问题,并与障碍函数法进行比较,避免障碍函数法由于罚因子趋于零导致的病态问题;最后给出一种非精确求解修正IPA的方法,并进行...文章对IPA(inerior point algorithm)进行了修正,证明修正的IPA的收敛性;将修正的IPA应用于不等式约束凸优化问题,并与障碍函数法进行比较,避免障碍函数法由于罚因子趋于零导致的病态问题;最后给出一种非精确求解修正IPA的方法,并进行数值实验,数值实验结果表明,算法是有效的.展开更多
This paper proposes an infeasible interior-point algorithm for linear complementarity problem with full-Newton steps.The main iteration consists of a feasibility step and several centrality steps.No more than O（n log...This paper proposes an infeasible interior-point algorithm for linear complementarity problem with full-Newton steps.The main iteration consists of a feasibility step and several centrality steps.No more than O（n log（n /ε））iterations are required for getting ε-solution of the problem at hand,which coincides with the best-known bound for infeasible interior-point algorithms.展开更多
In this paper a successive approximation method for solving probabilistic constrained programs is proposed. At each iteration of this method only few linear programs on a normal scale have to be solved. An error bound...In this paper a successive approximation method for solving probabilistic constrained programs is proposed. At each iteration of this method only few linear programs on a normal scale have to be solved. An error bound for the optimal value is given.展开更多
文摘文章对IPA(inerior point algorithm)进行了修正,证明修正的IPA的收敛性;将修正的IPA应用于不等式约束凸优化问题,并与障碍函数法进行比较,避免障碍函数法由于罚因子趋于零导致的病态问题;最后给出一种非精确求解修正IPA的方法,并进行数值实验,数值实验结果表明,算法是有效的.
基金Supported by the National Natural Science Foundation of China(71071119)
文摘This paper proposes an infeasible interior-point algorithm for linear complementarity problem with full-Newton steps.The main iteration consists of a feasibility step and several centrality steps.No more than O（n log（n /ε））iterations are required for getting ε-solution of the problem at hand,which coincides with the best-known bound for infeasible interior-point algorithms.
文摘In this paper a successive approximation method for solving probabilistic constrained programs is proposed. At each iteration of this method only few linear programs on a normal scale have to be solved. An error bound for the optimal value is given.
