何坤 郭洋俊骁 赵世莲
摘 要:最优性条件在优化问题中起着重要的作用,它为优化算法的研究提供了重要的理论依据。众所周知,凸规划方面最优性条件已比较完善。然而,由于拟凸函数性质的特殊性,对于拟凸规划问题解的Karush-Kuhn-Tucker(KKT)类型最优性条件的研究相对较少。本文利用半拟可微刻画了拟凸规划的最优性条件,同时研究了可行集法锥与带半拟可微性质的约束函数之间的关系,并证明了上述两个结果与Greenberg-Pierskalla次微分的关系。
关键词:半拟可微;次微分;拟凸规划;最优性条件;法锥
中图分类号:O224 文献标志码:A文章编号:1673-5072(2024)02-0150-05
拟凸函数及其性质的研究因其在数学、经济学、图像处理和机器学习等各个科学技术领域的应用而受到广泛关注[1-6]。在优化问题的研究中,最优性条件起着重要的作用。对于凸规划和拟凸规划问题,许多学者通过使用一些次微分,引入了各種类型的充分和必要最优性条件。然而,关于不可微拟凸规划的Karush-Kuhn-Tucker型(KKT型)最优性条件的结果并不多。
本文研究如下带不等式约束的拟凸规划问题:
minf(x),x∈K,(1)
近年来,在没有凸性的假设下,利用上正则凸化器逼近非凸函数得到非凸问题的最优性条件被广泛讨论。Kabgani[7]介绍了函数的半拟可微性质作为上正则凸化器的推广,并在拟凸的假设下用半拟可微刻画了函数的GP次微分。Suzuki[1]利用GP次微分证明了本质拟凸规划的充要KKT型最优性条件,但对于一般拟凸规划问题的KKT型最优性条件并没有研究,又因半拟可微性质良好,故想利用函数的半拟可微性质刻画问题(1)的KKT型最优性条件,同时研究问题(1)中可行集法锥与带半拟可微性质的约束函数之间的关系,最终形成一套完整的体系。
1 预备知识
2 一些引理
易知引理6—7成立:
3 主要结果
考虑问题(1),有以下定理:
证明 首先证明
参考文献:
[1] SUZUKI S.Karush-Kuhn-Tucker type optimality condition for quasiconvex programming in terms of Greenberg-Pierskalla subdifferential[J].Journal of Global Optimization,2021,79(1):191-202.
[2] AGRAWAL A,BOYD S.Disciplined quasiconvex programming[J].Optimization Letters,2020,14(7):1643-1657.
[3] HISHINUMA K,IIDUKA H.Fixed point quasiconvex subgradient method[J].European Journal of Operational Research,2020,282(2):428-437.
[4] PLASTRIA F.On the structure of the weakly efficient set for quasiconvex vector minimization[J].Journal of Optimization Theory and Applications,2020,184(2):547-564.
[5] SUZUKI S.Optimality conditions and constraint qualifications for quasiconvex programming[J].Journal of Optimization Theory and Applications,2019,183(3):963-976.
[6] ZHANG X,HE Z,ZHANG X,et al.High-performance beampattern synthesis via linear fractional semidefinite relaxation and quasi-convex optimization[J].IEEE Transactions on Antennas and Propagation,2018,66(7):3421-3431.
[7] KABGANI A.Characterization of nonsmooth quasiconvex functions and their Greenberg-Pierskallas subdifferentials using semi-quasidifferentiability notion[J].Journal of Optimization Theory and Applications,2021,189(2):666-678.
[8] GREENBERG H J,PIERSKALLA W P.Quasi-conjugate functions and surrogate duality[J].Cahiers du Centre Détude de Recherche Operationelle,1973,15:437-448.
[9]HIRIART-URRUTY J B,LEMARECHEL C.Convex analysis and minimization algorithms II[M].Heidelberg:Springer Berlin,1993.
[10]刘普寅.凸集支撑函数的性质及其应用[J].国防科技大学学报,1990(3):25-31.
[11]梅家騮.广义凸集的切锥[J].南昌大学学报(理科版),1990(4):24-30.
[12]ROCKAFELLAR R T.Convex Analysis[M].Princeton:Princeton University Press,1970.
[13]PENOT J P.Characterization of solution sets of quasiconvex programs[J].Journal of Optimization Theory and Applications,2003,117(3):627-636.
Karush-Kuhn-Tucker Type Optimality Conditionsfor Semi-quasi-differentiable Quasi-convex Programming
Abstract:As optimality condition plays an important role in the optimization problem,it provides an important theoretical basis for the study of optimization algorithm.It is well known that the optimality condition of convex programming has been relatively perfect.However,there are only few studies on Karush-Kuhn-Tucker type optimality conditions for the solutions of quasi-convex programming problems due to the special nature of quasi-convex functions.In this paper,the optimality conditions of quasi-convex programming are characterized by semi-quasi-differentiable,and the relationship between the feasible set normal cone and the constraint function with semi-quasi-differentiable properties is studied as well.In addition,the relationship between the above two results and Greenberg-Pierskalla subdifferential is proved.
Keywords:semi-quasi-differentiable;subdifferential;quasi-convex programming;optimality conditions;normal cone