DC复合优化问题的最优性条件
Optimality Conditions for DC Composite Optimization Problems with Conical Constraints
通讯作者:
收稿日期: 2020-09-29
基金资助: |
|
Received: 2020-09-29
Fund supported: |
|
作者简介 About authors
胡玲莉,E-mail:
田利萍,E-mail:
In the case when the functions are not necessarily lower semicontinuous and the sets are not necessarily closed, we first define the dual problem for DC composite optimization problems with conical constraints by using convexification technique, then some optimality conditions and saddle point theorems are obtained, which extend the corresponding results in the previous papers.
Keywords:
本文引用格式
胡玲莉, 田利萍, 方东辉.
Hu Lingli, Tian Liping, Fang Donghui.
1 引言
设
其中
建立了上述问题的弱对偶, 强对偶, 零对偶, 全对偶和最优性条件成立的充分和必要条件. 进一步, 文献[17]考虑了一类更具一般性的带复合函数的DC锥约束优化问题
其中
受上述文献的启发, 本文主要研究问题
2 预备知识
设
令
设
特别地, 由定义有
由文献[18, 定理2.3.1和定理2.4.2 (iii)] 知, Young-Fenchel不等式和Young等式成立, 即
进一步, 设
设函数
则称
显然,
设
由文献[19]的定理3.1可得
其中
引理2.1 令
3 复合优化问题的最优性条件与鞍点定理
设
设
定义3.1 设
则称系统
定理3.1 设
进一步, 若存在
证 设
故由(2.5)式可得
于是对任意的
从而由系统
进一步, 若存在
又由前面的证明可知
因此存在
下面考虑问题
设
则称
并对任意的
令
命题3.1 若
证 若
即
由Young-Fenchel不等式可知
注意到对任意的
从而
又
即
因此
定理3.2 设
(ⅰ)
(ⅱ)
(ⅲ)
(ⅳ)
证 设
(ⅰ)
由Young-Fenchel不等式可知
又
从而
因此(3.9)式的不等号变为等号, 即结论(ⅱ)成立.
(ⅱ)
又
(ⅱ)
因此(ⅱ)与(ⅲ)等价.
(ⅱ)
即结论(ⅳ)成立.
(ⅳ)
另一方面, 任取
因此结论(ⅲ)成立, 从而由(ⅱ)与(ⅲ)的等价性可得结论(ⅱ)成立. 证毕.
当
令
则由定理3.2可知以下推论成立.
推论3.1 设
(ⅰ)
(ⅱ)
(ⅲ)
(ⅳ)
例3.1 设
则
因此系统
因此定理3.1的结论成立. 进一步, 由共轭函数的定义可得
因此,
参考文献
On strong and total Lagrange duality for convex optimization problems
,
New regularity conditions for strong and total Fenchel-Lagrange duality in infinite dimensional spaces
,
Constraint qualifications for optimality conditions and total Lagrangian dualities in convex infinite programming
,
Complete characterization of stable Farkas' lemma and cone-convex programming duality
,
Optimality conditions for DC fractional programming problems
,
The stable Farkas lemma for composite convex functions in infinite dimensional spaces
,DOI:10.1007/s10255-015-0493-1 [本文引用: 2]
复合凸优化问题全对偶性的等价刻画
,
Some characterizations of total duality for a composed convex optimization
锥约束复合优化问题的Lagrange对偶
,DOI:10.12341/jssms13054 [本文引用: 1]
Study on the Lagrange dualities for composite optimization problems with conical constraints
DOI:10.12341/jssms13054 [本文引用: 1]
Approximate optimality conditions for composite convex optimization problems
,
带锥约束的复合优化问题的最优性条件
,DOI:10.3969/j.issn.1003-3998.2018.06.008 [本文引用: 2]
Optimality conditions for composite optimization problems with conical constraints
DOI:10.3969/j.issn.1003-3998.2018.06.008 [本文引用: 2]
Farkas-type results and duality for DC programs with convex constraints
,
Duality and Farkas-type results for DC infinite programming with inequality constraints
,
Some characterizations of duality for DC optimization with composite functions
,DOI:10.1080/02331934.2017.1338289 [本文引用: 1]
Zero duality gap properties for DC composite optimazation problem
,
Optimality conditions and total dualities for conic programming involving composite function
,DOI:10.1080/02331934.2018.1561695 [本文引用: 6]
Extended Farkas's lemmas and strong dualities for conic constraint problem involving composite functions
,DOI:10.1007/s10957-018-1219-3 [本文引用: 1]
Extended Farkas lemma and strong duality for composite optimization problems with DC functions
,DOI:10.1080/02331934.2016.1266628 [本文引用: 2]
Frchet subdifferential calculus and optimality conditions in nondifferentiable programming
,DOI:10.1080/02331930600816395 [本文引用: 1]
/
〈 | 〉 |