等式约束最优化问题MBFGS法的全局收敛性  

On Global Convergence of the MBFGS Method in Equality Constrained Optimization

作　　者：蒋莉[1] 

机构地区：[1]湖南农业大学理学院,湖南长沙410128

出　　处：《湖南农业大学学报（自然科学版）》2006年第3期324-326,共3页Journal of Hunan Agricultural University(Natural Sciences)

基　　金：湖南农业大学人才科学基金(03YJ14)

摘　　要：利用Li-Fukushima提出的求解无约束问题的修正BFGS(MBFGS)公式,提出了求解等式约束问题的SQP算法,并利用l1精确罚函数进一步将算法全局化,证明了在一定条件下算法的全局收敛性.此方法的最大特点是能够保证SQP算法子问题中矩阵的对称正定性.Based on a modified BFGS update formula proposed by Li-Fukushima, a modified BFGS method for solving equality constrained optimization was proposed. The remarkable feature of this method is that the matrices in SQP＇s subproblems could be kept to be positivly definite. Based on this algorithm which possesses a local convergence property, the proposed SQP algorithm for solving equality constrained optimization by using the li exact penalty function was globalized. Under appropriate conditions, the global convergence of the globalized method was obtained.

关 键 词：BFGS修正 全局收敛性 SQP算法 等式约束问题 

分 类 号：O212.2[理学—概率论与数理统计]