基于积极集识别技术的半无限minimax问题非单调有限记忆SQP算法

A NON-MONOTONE SQP ALGORITHM WITH L-BFGS UPDATE FOR SOLVING SEMI-INFINITE MINIMAX PROBLEM BASED ON ACTIVE SET TECHNIQUE

本文研究了半无限minimax问题.利用积极集识别技术结合非单调有限记忆序列二次规划(SQP)方法来求解半无限minimax问题.在适当的条件下证明了算法的收敛性.数值结果表明新算法在降低求解规模和迭代次数等方面均优于采用Armijo型线搜索的SQP方法.

This paper studies the semi-infinite minimax problem.By using active set recognition technology combined with non-monotonic finite memory sequential quadratic programming(SQP) method to solve the semi-infinite minimax problem.The convergence of the algorithm is proved under appropriate conditions.Numerical results show that the new algorithm is superior to the SQP method using Armijo-type line search in terms of reducing the solution scale and the number of iterations.