摘要
将求解半无限规划离散化问题的一个可行模松弛SQP算法推广到离散的半无限极大极小问题,提出一个全局收敛的模松弛SQP算法.算法要求迭代点可行,且每次迭代只需求解一个二次规划(QP)子问题即可获得搜索方向.通过修正其离散指标集,使得每次迭代求解QP子问题时只需利用一小部分离散指标即可,这大大降低了计算成本.在合适的条件下,可证明算法具有全局收敛性.
In this paper. we extend the feasible norm.,relaxed SQP algorithm (Jian. Xu and ltan, 2008) for dis- aretizcd scmi-inlinitc problems to the discrctizcd semi-infinite minimax i)rohlems, and present a globally convergent nornl-rtlaxtd SQP algorithm. At each iteratim, the iteration poinl is flsit, and m imprmc,d direction is ol)ttind by solving only one quadratic Irogramming (QP) SUblrolem. Only a few of discrttizt.d indicesm:, used in the, QP subprobhms bylid;ting lhc di.creized index ses, which cm rduce the cotrl)tztalioz-1 largtdy. Und,r some ;ploro prialc conditions, the glolml convergence is Drovcd.
出处
《广西师范学院学报(自然科学版)》
2013年第2期1-7,共7页
Journal of Guangxi Teachers Education University(Natural Science Edition)
基金
supported by the Guangxi Education Office Research Project(201106LX322)
the Guangxi Teachers Education University Pilot Research Project(the 2010 project)