不等式约束Minimax问题的一个可行序列线性方程组算法

A Feasible Sequential Systems of Linear Equations Algorithm for Inequality Constrained Minimax Problem

提出了求解不等式约束minimax问题的一个可行序列线性方程组算法.在每次迭代中可行下降方向通过求解两个具有相同系数矩阵的线性方程组产生,系数矩阵具有较好的稀疏性,计算量少.在适当的假设条件下,算法具有全局收敛性,初步数值实验验证了算法的有效性.

A feasible sequential system of linear equations algorithm for inequality con- strained minimax problems is proposed. At each iteration,the proposed algorithm solves only two systems of linear equations with a same coefficient matrix obtain the feasible descent direction. Furthermore,the sparsity of the coefficient matrix is good. It is proved that under appropriate as- sumptions, the algorithm is globally convergent. Finally, several numerical examples are reported to verify effectiveness of the algorithm.