AAA Abstract and Applied Analysis 1687-0409 1085-3375 Hindawi Publishing Corporation 317304 10.1155/2014/317304 317304 Research Article On Nonsmooth Semi-Infinite Minimax Programming Problem with ( Φ , ρ ) -Invexity Liu X. L. 1 Lai G. M. 2 Xu C. Q. 3 Yuan D. H. 1 Pečarić Josip E. 1 Department of Mathematics Hanshan Normal University Guangdong 521041 China hstc.edu.cn 2 Department of Computer Science Hanshan Normal University Guangdong 521041 China hstc.edu.cn 3 Science College Beijing University of Civil Engineering and Architecture Beijing 102616 China bucea.edu.cn 2014 432014 2014 29 08 2013 12 01 2014 04 03 2014 2014 Copyright © 2014 X. L. Liu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We are interested in a nonsmooth minimax programming Problem (SIP). Firstly, we establish the necessary optimality conditions theorems for Problem (SIP) when using the well-known Caratheodory's theorem. Under the Lipschitz ( Φ , ρ ) -invexity assumptions, we derive the sufficiency of the necessary optimality conditions for the same problem. We also formulate dual and establish weak, strong, and strict converse duality theorems for Problem (SIP) and its dual. These results extend several known results to a wider class of problems.

1. Introduction

Convexity plays a central role in many aspects of mathematical programming including analysis of stability, sufficient optimality conditions, and duality. Based on convexity assumptions, nonlinear programming problems can be solved efficiently. There have been many attempts to weaken the convexity assumptions in order to treat many practical problems. Therefore, many concepts of generalized convex functions have been introduced and applied to mathematical programming problems in the literature . One of these concepts, invexity, was introduced by Hanson in . Hanson has shown that invexity has a common property in mathematical programming with convexity that Karush-Kuhn-Tucker conditions are sufficient for global optimality of nonlinear programming under the invexity assumptions. Ben-Israel and Mond  introduced the concept of preinvex functions which is a special case of invexity. Many other concepts of generalized convexity such as ( p , r ) -invexity , ( F , ρ ) -convexity , ( F , α , ρ , d ) -convexity , ( C , α , ρ , d ) -convexity , and V - r -invexity  have also been introduced. With these definitions of generalized invexity on the hand, several authors have been interested recently in the optimality conditions and duality results for different classes of minimax programming problems; see  for details.

Recently, Antczak and Stasiak  generalized the definition of ( Φ , ρ ) -invexity notion introduced by Caristi et al. and M. V. Ştefănescu and A. Ştefănescu [14, 15] for differentiable optimization problems to the case of mathematical programming problems with locally Lipschitz functions. They proved sufficient optimality conditions and duality results for nondifferentiable optimization problems involving locally Lipschitz ( Φ , ρ ) -invex functions. Antczak  also considered a class of nonsmooth minimax programming problems in which functions involved are locally Lipschitz ( Φ , ρ ) -invex. We point out that this locally Lipschitz ( Φ , ρ ) -invexity includes the ( C , α , ρ , d ) -convexity as a special case and Yuan et al.  defined firstly the ( C , α , ρ , d ) -convexity with a convex functional C .

Due to a growing number of theoretical and practical applications, semi-infinite programming has recently become one of the most substantial research areas in applied mathematics and operations research. For more details on semi-infinite programming we refer to the survey papers  and for clear understanding of different aspects of semi-infinite programming we refer to . M. V. Ştefănescu and A. Ştefănescu  considered differentiable Problem ( SIP ) with new ( Φ , ρ ) -invexity. However, the results of this kind of programming can not be used to deal with the concrete nonsmooth semi-infinite minimax programming problem as presenting in Example 9 in Section 3 since the objective function is nondifferentiable at x = 1 . Therefore, we are interested in dealing with nonsmooth Problem ( SIP ) with locally Lipschitz ( Φ , ρ ) -invexity proposed in , in this paper.

The rest of the paper is organized as follows. In Section 2, we present concepts regarding Lipschtiz ( Φ , ρ ) -invexity. In Section 3, we present not only necessary but also sufficient optimality conditions for nonsmooth Problem ( SIP ) . When the necessary optimality conditions and the ( Φ , ρ ) -invexity concept are utilized, dual Problem ( DI ) is formulated for the primal ( SIP ) and duality results between them are presented in Section 4. Section 5 is our conclusions.

2. Notations and Preliminaries

In this section, we provide some definitions and results that we shall use in the sequel. Let X be a subset of n and denote Q : = { 1,2 , , q } , Q * : = { 1,2 , , q * } , M : = { 1,2 , , m } , and M * : = { 1,2 , , m * } .

Definition 1.

A real-valued function f : X is said to be locally Lipschitz on X if, for any x X , there exist a neighborhood U of x and a positive constant T x > 0 such that (1) | f ( y ) - f ( z ) | T x y - z , y , z U .

Definition 2 (see [<xref ref-type="bibr" rid="B21">21</xref>]).

Let d n and f : X . If (2) f 0 ( x ; d ) = lim y x μ 0 sup 1    μ    ( f ( y + μ d ) - f ( y ) ) exists, then f 0 ( x ; d ) is said to be the Clarke derivative of f at x in the direction d . If this limit superior exists for all d n , then f is called Clarke differentiable at x . The set (3) f ( x ) = { ζ f 0 ( x ; d ) ζ , d , d n } is called the Clarke subgradient of f at x .

Note that if a function is locally Lipschitz, then its Clarke subgradient must exist.

The definition of the locally Lipschitz ( Φ , ρ ) -invexity was introduced by Antczak and Stasiak ; see also the following Definition 3. This generalized invexity was introduced as a generalization of differentiable ( Φ , ρ ) -invexity notion defined by Caristi et al. and M. V. Ştefănescu and A. Ştefănescu in [14, 17]. The main tool used in the definition of the locally Lipschitz ( Φ , ρ ) -invexity notion is the above Clarke generalized subgradient (see Definition 2).

Definition 3.

Let f : X be a real-valued Lipschitz function on X . For fixed u X , let Φ : X × X × n + 1 be convex with respect to the third argument on n + 1 such that Φ ( x , u , ( 0 , a ) ) 0 for every x X and any a 0 . If there exists a real-valued function ρ ( · , · ) : X × X such that (4) f ( x ) - f ( u ) ( > ) Φ ( x , u , ( ξ , ρ ( x , u ) ) ) , ξ f ( u ) holds for all x X ( x u ), then f is said to be (strictly) locally Lipschitz ( Φ , ρ ) -invex at u on X or shortly (strictly) ( Φ , ρ ) -invex at u on X . If f is (strictly) locally Lipschitz ( Φ , ρ ) -invex at any u of X , then f is (strictly) locally Lipschitz ( Φ , ρ ) -invex on X .

Remark 4.

In order to define an analogous class of (strictly) locally Lipschitz ( Φ , ρ ) -incave functions, the direction of the inequality in the definition of these functions should be changed to the opposite one.

In this paper, we deal with the nonsmooth semi-infinite minimax programming Problem ( SIP ) with the locally Lipschitz ( Φ , ρ ) -invexity proposed by Antczak and Stasiak . Here, Problem ( SIP ) is ( SIP ) min sup y Y sup y Y ϕ ( x , y ) subject to ψ ( x , z ) 0 , z Z , where Y and Z are compact subsets of some Hausdorff topological spaces, ϕ ( · , · ) : n × Y , ψ ( · , · ) : n × Z . Let E SIP be the set of feasible solutions of Problem ( SIP ) ; in other words, E SIP = { x n ψ ( x , z ) 0 , z Z } . For convenience, let us define the following sets for every x E SIP : (5) J ( x ) = { z Z ψ ( x , z ) = 0 } , Z ( x ) = { z * Z ψ ( x , z * ) = sup z Z ψ ( x , z ) } , Y ( x ) = { y * Y ϕ ( x , y * ) = sup y Y ϕ ( x , y ) } .

If x E SIP , then J ( x ) represents the index set of the active restrictions at x . Note that J ( x ) = Z ( x ) when J ( x ) is not empty.

Consider the nonlinear programming problem ( P ) min f ( x ) subject    to g ( x ) 0 , where f , g : n . A particular case of Problem ( P ) is the minimax problem ( SIP ) in which the functions f , g are given by (6) f ( x ) = sup y Y ϕ ( x , y ) , (7) g ( x ) = sup z Z ψ ( x , z ) , respectively. Let x 0 E SIP . Consider the following unconstrained optimization problem ( P ) : ( P ) min { h ( x ) x n } , where h ( x ) = max { f ( x ) - f ( x 0 ) , g ( x ) } . Then, the relationship between Problems ( P ) and ( P ) is given in the following lemma.

Lemma 5.

If x 0 is a local minimizer for Problem ( P ) , then x 0 is also a local minimizer for Problem ( P ) .

To deal with the nonsmooth Problem ( SIP ) , we need the following Conditions 1 and 2.

Condition 1.

We assume that (a) the sets Y and Z are compact; (b) the function ϕ ( x , y ) is upper semicontinuous in ( x , y ) , and the function ψ ( x , z ) is upper semicontinuous in ( x , z ) ; (c) the function ϕ ( x , y ) is locally Lipschitz in x and uniformly for y in Y , and the function ψ ( x , y ) is locally Lipschitz in x and uniformly for z in Z ; (d) the function ϕ ( x , y ) is regular in x ; that is, ϕ x ( x , y ; · ) = ϕ x ( x , y ; · ) , where the symbol ϕ x denotes the derivative with respect to x ; also the function ψ ( x , z ) is regular in x ; (e) the set-valued map ϕ x ( x , y ) is upper semicontinuous in ( x , y ) , and the set-valued map ψ x ( x , z ) is upper semicontinuous in ( x , z ) .

Condition 2.

For any finite subset Z Z ( x * ) , for some x * E SIP , the equality z Z η z ζ z = 0 with η z 0 , ζ z ψ ( x * , z ) , z Z , implies that (8) η z = 0 , z Z .

Clarke [22, Theorem 2.1] has shown that, under the assumptions (a)–(e) of Condition 1, the maximum function f defined by (6) is locally Lipschitz; f ( x , d ) exists and is given by the formula (9) f ( x , d ) = f ( x , d ) = max { ξ · d ξ x ϕ ( x , y ) , y Y ( x ) } , where ξ · d denotes the inner product of vectors ξ and d . Moreover, the sup in (6) can be replaced by max and the subgradient f ( x ) is given by (10) f ( x ) = co { y Y ( x ) x ϕ ( x , y ) } .

Similarly, the maximum function g defined by (7) is locally Lipschitz; g ( x , d ) and g ( x ) are given by (11) g ( x , d ) = g ( x , d ) = max { ζ · d ζ x ψ ( x , z ) , z Z ( x ) } , g ( x ) = co { z Z ( x ) x ψ ( x , z ) } , respectively.

3. Optimality Conditions

In this section, we establish not only the necessary optimality conditions theorems but also the sufficient optimality conditions theorems for Problem ( SIP ) with the functions involved being locally Lipschitz with respect to the variable x .

Theorem 6 (necessary optimality conditions).

Let x * be an optimal solution of ( SIP ) . One also assume that Condition 1 holds. Then there exist nonnegative integers q * and m * with 1 q * + m * n + 1 , vectors y i * Y ( x * ) ( i Q * ), z j * Z ( x * ) ( j M * ), and scalars λ i * > 0 ( i Q * ), μ j * > 0 ( j M * ) such that (12) 0 i = 1 q * λ i * x ϕ ( x * , y i * ) + j = 1 m * μ j * x ψ ( x * , z j * ) , (13) i = 1 q * λ i * + j = 1 m * μ j * 0 .

Here, one allow the case, where if q * = 0 , then the set Q * is empty; similarly, if m * = 0 , then the set M * is empty.

Proof.

Let x * be a local minimizer for Problem ( SIP ) . This means that x * is a local minimizer for Problem ( P ) , where f and g are given by (6) and (7), respectively. Therefore, x * is a local minimizer for Problem ( P ) .

By Condition 1 and [22, Theorem 2.1], f is locally Lipschitzian and regular at x * , so the function (14) f ~ ( x ) = f ( x ) - f ( x * ) has the same properties. Then, using [21, Propositions 2.3 . 2 and 2.3 . 12 ], we obtain (15) 0 h ( x * ) = co { f ~ ( x * ) g ( x * ) } = co { f ( x * ) g ( x * ) } = co { co { y Y ( x * ) x ϕ ( x * , y ) } co { z Z ( x * ) x ψ ( x * , z ) } } = co { y Y ( x * ) x ϕ ( x * , y ) z Z ( x * ) x ψ ( x * , z ) } ; here the equality (16) co ( co A co B ) = co ( A B ) is used in the fourth equality. Hence, by Caratheodory's theorem, there exist the nonnegative integers q * and m * and the scalars λ i * > 0 ( i Q * ) and μ j * > 0 ( j M * ) such that (17) 1 q * + m * n + 1 , 0 = i = 1 q * λ i ξ i + j = 1 m * μ j ζ j for some ξ i y Y ( x * ) x ϕ ( x * , y ) and ζ j z Z ( x * ) x ψ ( x * , z ) . Note that, for each i Q * , ξ i y Y ( x * ) x ϕ ( x * , y ) means that there exists y Y ( x * ) such that ξ i x ϕ ( x * , y ) , and denote this y by y i * . Similarly, there exists z j * Z ( x * ) such that ζ j x ψ ( x * , z j * ) for each j M * . Now the desired inequalities (12) and (13) can be deduced from the above discussion.

Theorem 7 (necessary optimality conditions).

Let x * be an optimal solution of ( SIP ) . One also assume that Conditions 1 and 2 hold. Then there exist the nonnegative integers q * > 0 and m * with 1 q * + m * n + 1 , the vectors y i * Y ( x * ) ( i Q * ), z j * Z ( x * ) ( j M * ), and the scalars λ i * > 0 ( i Q * ), μ j * > 0 ( j M * ) satisfying (12) and (18) i = 1 q * λ i * = 1 .

Proof.

By Theorem 6, we need to prove q * 0 , on the contrary, that is, q * = 0 , then one obtains from (12) and (13) that (19) 0 j = 1 m * μ j * x ψ ( x * , z j * ) , (20) j = 1 m * μ j * 0 , respectively. By (19), there exist ζ j x ψ ( x * , z j * ) for j M * satisfying (21) j = 1 m * μ j * ζ j = 0 .

Now one obtains from the assumptions of Condition 2 that μ j * = 0 for j M * ; this contradicts to (20), and we obtain the desired results.

Next, we derive a sufficient optimality conditions theorem for Problem ( SIP ) under the assumption of ( Φ , ρ ) -invexity as defined in Definition 3.

Theorem 8 (sufficient optimality conditions).

Let ( x * , q * , m * , λ * , μ * , y * ¯ , z * ¯ ) satisfy conditions (12) and (18), where y * ¯ = ( y 1 * , , y q * * ) and z * ¯ = ( z 1 * , , z m * * ) . Assume that ϕ ( · , y i * ) , i Q * , are ( Φ , ρ i ϕ ) -invex at x * on X and ϕ ( · , z j * ) , j M * , are ( Φ , ρ j ψ ) -invex at x * on X . If J ( x * ) and (22) i = 1 q * λ i * ρ i ϕ ( x , x * ) + j = 1 m * μ j * ρ j ψ ( x , x * ) 0 , x X , then x * is an optimal solution to ( SIP ) .

Proof.

Suppose, contrary to the result, that x * is not an optimal solution for Problem ( SIP ) . Hence, there exists x 0 E SIP such that (23) sup y Y ϕ ( x 0 , y ) < ϕ ( x * , y i * ) , i Q * , ψ ( x 0 , z ) 0 = ψ ( x * , z j * ) , j M * .

Thus, (24) ϕ ( x 0 , y i * ) < ϕ ( x * , y i * ) , i Q * , ψ ( x 0 , z j * ) ψ ( x * , z j * ) , j M * .

Now, we can write the following statement: (25) i = 1 q * λ i * ( ϕ ( x 0 , y i * ) - ϕ ( x * , y i * ) ) + j = 1 m * μ j * ( ψ ( x 0 , z j * ) - ψ ( x * , z j * ) ) < 0 .

By the generalized invexity assumptions of ϕ ( · , y i * ) and ψ ( · , z j * ) , we have (26) ϕ ( x 0 , y i * ) - ϕ ( x * , y i * ) Φ ( x 0 , x * , ( ξ i , ρ i ϕ ( x 0 , x * ) ) ) , ξ i x ϕ ( x * , y i * ) , ψ ( x 0 , z j * ) - ψ ( x * , z j * ) Φ ( x 0 , x * , ( ζ j , ρ j ψ ( x 0 , x * ) ) ) , ζ j x ψ ( x * , z j * ) .

Employing (26) to (25), we have (27) i = 1 q * λ i * Φ ( x 0 , x * , ( ξ i , ρ i ϕ ( x 0 , x * ) ) ) + j = 1 m * μ j * Φ ( x 0 , x * , ( ζ j , ρ j ψ ( x 0 , x * ) ) ) < 0 .

By (18) and the convexity of Φ , we deduce that (28) Φ ( x 0 , x * , ( i = 1 q * λ i * ξ i + j = 1 m * μ j * ζ j , i = 1 q * λ i * ρ i ϕ ( x 0 , x * ) + j = 1 m * μ j * ρ j ψ ( x 0 , x * ) ) ) < 0 .

This, together with (40) and the assumption Φ ( x 0 , x * , ( 0 , a ) ) 0 for any a 0 , follows that (29) 0 i = 1 q * λ i * x ϕ ( x * , y i * ) + j = 1 m * μ j * x ψ ( x * , z j * ) .

This is a contradiction to condition (12).

Example 9.

Let Y = [ 0,1 ] and Z = [ 1,2 ] . Define (30) ϕ ( x , y ) = { x + 2 y , y x , 2 x + y , x < y , ψ ( x , z ) = | x - 3 2 | - z 2 .

Then, ϕ ( · , y ) is ( Φ , 1 ) -invex at x = 1 for each y Y , ψ is ( Φ , 1 ) -invex at x = 1 for each z Z , and (31) f ( x ) = sup y Y ϕ ( x , y ) = { x + 2 , y x , 2 x + 1 , x < y , g ( x ) sup z Z ψ ( x , z ) = | x - 3 2 | - 1 2 .

Note that (32) x ϕ ( x , y ) = { 1 , y x , [ 1,2 ] , x = y , 2 , x < y , x ψ ( x , z ) = x ψ ( x , z ) = 1 .

Consider x 0 = 1 . Since Y ( x 0 ) = { 1 } , then we can assume q = 1 . Therefore, (33) 0 λ x ϕ ( 1,1 ) - μ x ψ ( 1,1 ) = λ [ 1,2 ] - μ , where λ = μ = 1 . Now, from Theorem 8, we can say that x 0 = 1 is an optimal solution to ( SIP ) .

4. Duality

Making use of the optimality conditions of the preceding section, we present dual Problem ( DI ) to the primal one ( SIP ) and establish weak, strong, and strict converse duality theorems. For convenience, we use the following notations: (34) K ( x ) = { with z j Z ( x ) , j M i = 1 q ( q , m , λ , y ¯ , z ¯ ) 1 q q + m n + 1 , λ = ( λ 1 , , λ q ) with i = 1 q λ i = 1 , λ i > 0 for i Q , y ¯ = ( y 1 , , y q ) with y i Y ( x ) , i Q , z ¯ = ( z 1 , , z m ) with z j Z ( x ) , j M i = 1 q } .

Our dual problem ( DI ) can be stated as follows: ( DI ) max ( q , m , λ , y ¯ , z ¯ ) K ( w ) sup ( w , μ , ν ) H 1 ( q , m , λ , y ¯ , z ¯ ) ν , where H 1 ( q , m , λ , y ¯ , z ¯ ) denotes the set of all ( w , μ , ν ) satisfying (35) 0 i = 1 q λ i x ϕ ( w , y i ) + j = 1 m μ j x ψ ( w , z j ) , w X , (36) ϕ ( w , y i ) ν 0 ,    i Q , (37) μ j ψ ( w , z j ) 0 , j M , (38) μ = ( μ 1 , , μ m ) with μ j > 0 for j M , (39) ( q , m , λ , y ¯ , z ¯ ) K ( w ) .

Note that if H 1 ( q , m , λ , y ¯ , z ¯ ) is empty for some ( q , m , λ , y ¯ , z ¯ ) K ( w ) , then define sup ( w , μ , ν ) H 1 ( q , m , λ , y ¯ , z ¯ ) ν = - .

Theorem 10 (weak duality).

Let x and ( w , q , m , λ , μ , ν , y ¯ , z ¯ ) be ( SIP ) -feasible and ( DI ) -feasible, respectively. Suppose that ϕ ( · , y i ) , i Q , are ( Φ , ρ i ϕ ) -invex at w on X and ψ ( · , z j ) , j M , are ( Φ , ρ j ψ ) -invex at w on X . If (40) i = 1 q λ i ρ i ϕ ( x , w ) + j = 1 m μ j ρ j ψ ( x , w ) 0 , then (41) sup y Y ϕ ( x , y ) ν .

Proof.

Suppose to the contrary that sup y Y ϕ ( x , y ) < ν . Therefore, we obtain (42) ϕ ( x , y ) < ν ϕ ( w , y i ) , y i Y ( w ) , y Y , i Q .

Thus, we obtain (43) ϕ ( x , y i ) < ϕ ( w , y i ) , y i Y ( w ) , i Q .

Note that (44) ψ ( x , z ) 0 , z Z , μ j ψ ( w , z j ) 0 , μ j > 0 , z j Z ( w ) , j M .

We obtain that (45) μ j ψ ( x , z j ) μ j ψ ( w , z j ) , j M .

Therefore, (46) i = 1 q λ i ( ϕ ( x , y i ) - ϕ ( w , y i ) ) + j = 1 m μ j ( ψ ( x , z j ) - ψ ( w , z j ) ) < 0 .

Similar to the proof of Theorem 8, by (46) and the generalized invexity assumptions of ϕ ( · , y i ) and ψ ( · , z j ) , we have (47) Φ ( x , w , ( j = 1 m i = 1 q λ i ξ i + j = 1 m μ j ζ j , i = 1 q λ i ρ i ϕ ( x , w ) + j = 1 m μ j ρ j ψ ( x , w ) ) ) < 0 .

This follows that (48) 0 i = 1 q λ i x ϕ ( w , y i ) + j = 1 m μ j x ψ ( w , z j ) .

Thus, we have a contradiction to (35). So sup y Y ϕ ( x , y ) ν .

Theorem 11 (strong duality).

Let Problem ( SIP ) satisfy Conditions 1 and 2; let x * be an optimal solution of Problem ( SIP ) . If the hypothesis of Theorem 10 holds for all ( DI ) -feasible points ( w , q , m , λ , μ , ν , y ¯ , z ¯ ) , then there exists ( q * , m * , λ * , y * ¯ , z * ¯ ) K ( x * ) , ( x * , μ * , ν * ) H 1 ( q * , m * , λ * , y * ¯ , z * ¯ ) such that ( x * , q * , m * , λ * , μ * , ν * , y * ¯ , z * ¯ ) is a ( DI ) -optimal solution, and the two problems ( P ) and ( DI ) have the same optimal values.

Proof.

By Theorem 7, there exists ν * = ϕ ( x * , y i * ) ( i = 1 , , q * ), satisfying the requirements specified in the theorem, such that ( x * , q * , m * , λ * , μ * , ν * , y * ¯ , z * ¯ ) is a ( DI ) -feasible solution; then the optimality of this feasible solution for ( DI ) follows from Theorem 10.

Theorem 12 (strict converse duality).

Let x 0 and ( w , q , m , λ , μ , ν , y ¯ , z ¯ ) be optimal solutions of ( SIP ) and ( DI ) , respectively. Suppose that ϕ ( · , y i ) are strictly ( Φ , ρ i ϕ ) -invex at w for each i Q and ψ ( · , z j ) are ( Φ , ρ i ψ ) -invex at w for each j M . If (49) i = 1 q λ i ρ i ϕ ( x 0 , w ) + j = 1 m μ j ρ j ψ ( x 0 , w ) 0 , then x 0 = w ; that is, w is a ( SIP ) -optimal solution, and (50) sup y Y ϕ ( x 0 , y ) = ν .

Proof.

Suppose to the contrary that x 0 w . By the generalized invexity assumptions of ϕ ( · , y i ) and ψ ( · , z j ) , we have (51) ϕ ( x 0 , y i ) - ϕ ( w , y i ) > Φ ( x 0 , w , ( ξ i , ρ i ϕ ( x 0 , w ) ) ) , ξ i x ϕ ( w , y i ) , ψ ( x 0 , z j ) - ψ ( w , z j ) Φ ( x 0 , w , ( ζ j , ρ j ψ ( x 0 , w ) ) ) , ζ j x ψ ( w , z j ) .

Therefore, we obtain from (51) and the convexity of Φ that (52) i = 1 q λ i ( ϕ ( x 0 , y i ) - ϕ ( w , y i ) ) + j = 1 m μ j ( ψ ( x 0 , z j ) - ψ ( w , z j ) ) > Φ ( x 0 , w , ( i = 1 q λ i ξ i + j = 1 m μ j ζ j , i = 1 q λ i ρ i ϕ ( x 0 , w ) + j = 1 m μ j ρ j ψ ( x 0 , w ) ) ) holds for all ξ i x ϕ ( w , y i ) and ζ j x ψ ( w , z j ) . This, together with (35), (49), and (53) Φ ( x 0 , w , ( 0 , a ) ) 0 for a > 0 , follows that (54) i = 1 q λ i ( ϕ ( x 0 , y i ) - ϕ ( w , y i ) ) + j = 1 m μ j ( ψ ( x 0 , z j ) - ψ ( w , z j ) ) > 0 while (55) j = 1 m μ j ( ψ ( x 0 , z j ) - ψ ( w , z j ) ) 0 .

Thus (56) i = 1 q λ i ( ϕ ( x 0 , y i ) - ϕ ( w , y i ) ) > 0 .

From the above inequality, we can conclude that there exists i 0 Q , such that (57) ϕ ( x 0 , y i 0 ) - ϕ ( w , y i 0 ) > 0 or (58) ϕ ( x 0 , y i 0 ) > ϕ ( w , y i 0 ) .

It follows that (59) sup y Y ϕ ( x 0 , y ) ϕ ( x 0 , y i 0 ) > ϕ ( w , y i 0 ) > ν .

On the other hand, we know from Theorem 10 that (60) sup y Y ϕ ( x 0 , y ) = ν .

5. Conclusions

In this paper, we have discussed a nonsmooth semi-infinite minimax programming Problem ( SIP ) . We have extended the necessary optimality conditions for Problem ( SIP ) considered in  to the nonsmooth case; we have also extended the sufficient optimality conditions and dual results of Problem ( SIP ) addressed by M. V. Ştefănescu and A. Ştefănescu in  to the nonsmooth case under the Lipschitz ( Φ , ρ ) -invexity assumptions as defined in . More exactly, we have established the necessary optimality conditions theorems for the Problem ( SIP ) when using Caratheodory's theorem. Under the Lipschitz ( Φ , ρ ) -invexity assumptions as defined in , we have derived the sufficiency of the necessary optimality conditions for Problem ( SIP ) . In the end, we have constructed a dual model ( DI ) and derived duality results between Problems ( SIP ) and ( DI ) . These results extend several known results to a wider class of problems.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

This research is supported by the Natural Science Foundation of Guangdong Province (Grant no. S2013010013101) and the Foundation of Hanshan Normal University (Grant nos. QD20131101 and LQ200905).

Antczak T. Generalized fractional minimax programming with B - ( p , r ) -invexity Computers & Mathematics with Applications 2008 56 6 1505 1525 10.1016/j.camwa.2008.02.039 MR2440568 Hanson M. A. On sufficiency of the Kuhn-Tucker conditions Journal of Mathematical Analysis and Applications 1981 80 2 545 550 10.1016/0022-247X(81)90123-2 MR614849 ZBL0463.90080 Ben-Israel A. Mond B. What is invexity? The Journal of the Australian Mathematical Society B 1986 28 1 1 9 10.1017/S0334270000005142 MR846778 ZBL0603.90119 Antczak T. ( p , r ) -invex sets and functions Journal of Mathematical Analysis and Applications 2001 80 545 550 10.1006/jmaa.2001.7574 MR1866053 ZBL1051.90018 Preda V. On Sufficiency and duality for multiobjective programs Journal of Mathematical Analysis and Applications 1992 166 2 365 377 10.1016/0022-247X(92)90303-U MR1160932 ZBL0764.90074 Liang Z. A. Huang H. X. Pardalos P. M. Optimality conditions and duality for a class of nonlinear fractional programming problems Journal of Optimization Theory and Applications 2001 110 3 611 619 10.1023/A:1017540412396 MR1854018 ZBL1064.90047 Yuan D. H. Liu X. L. Chinchuluun A. Pardalos P. M. Nondifferentiable minimax fractional programming problems with ( C , α , ρ , d ) -convexity Journal of Optimization Theory and Applications 2006 129 1 185 199 10.1007/s10957-006-9052-5 MR2281051 Antczak T. The notion of V - r -invexity in differentiable multiobjective programming Journal of Applied Analysis 2005 11 1 63 79 10.1515/JAA.2005.63 MR2151394 Lai H. C. Lee J. C. On duality theorems for a nondifferentiable minimax fractional programming Journal of Computational and Applied Mathematics 2002 146 1 115 126 10.1016/S0377-0427(02)00422-3 MR1923450 ZBL1010.90080 Antczak T. Minimax programming under ( p , r ) -invexity European Journal of Operational Research 2004 158 1 1 19 10.1016/S0377-2217(03)00352-7 MR2063607 Ahmad I. Gupta S. K. Kailey N. Agarwal R. P. Duality in nondifferentiable minimax fractional programming with B - ( p , r ) -invexity Journal of Inequalities and Applications 2011 2011 article 75 10.1186/1029-242X-2011-75 MR2847586 Mishra S. K. Shukla K. Nonsmooth minimax programming problems with V - r -invex functions Optimization 2010 59 1 95 103 10.1080/02331930903500308 MR2765470 Antczak T. Stasiak A. ( Φ , ρ ) -invexity in nonsmooth optimization Numerical Functional Analysis and Optimization 2011 32 1 1 25 10.1080/01630563.2010.526408 MR2739376 ZBL1229.90133 Caristi G. Ferrara M. Ştefănescu A. Konnov Igor V. Dinh T. L. Rubinov Alexander M. Mathematical programming with ( Φ , ρ ) -invexity Generalized Convexity and Related Topics 2007 583 Berlin, Germany Springer 167 176 Lecture Notes in Economics and Mathematical Systems 10.1007/978-3-540-37007-9_9 MR2279401 Ştefănescu M. V. Ştefănescu A. Minimax programming under new invexity assumptions Revue Roumaine de Mathématiques Pures et Appliquées 2007 52 3 367 376 MR2362856 ZBL1150.90446 Antczak T. Nonsmooth minimax programming under locally Lipschitz ( Φ , ρ ) -invexity Applied Mathematics and Computation 2011 217 23 9606 9624 10.1016/j.amc.2011.04.037 MR2811236 Ştefănescu M. V. Ştefănescu A. On semi-infinite minmax programming with generalized invexity Optimization 2012 61 11 1307 1319 10.1080/02331934.2011.563304 MR2995499 ZBL06115650 López M. Still G. Semi-infinite programming European Journal of Operational Research 2007 180 2 491 518 10.1016/j.ejor.2006.08.045 MR2295358 ZBL1124.90042 Shapiro A. Semi-infinite programming, duality, discretization and optimality conditions Optimization 2009 58 2 133 161 10.1080/02331930902730070 MR2501606 ZBL1158.90410 Floudas C. A. Pardalos P. M. Encyclopedia of Optimization 2009 2nd New York, NY, USA Springer Clarke F. H. Optimization and Nonsmooth Analysis 1983 New York, NY, USA John Wiley & Sons MR709590 Clarke F. H. Generalized gradients and applications Transactions of the American Mathematical Society 1975 205 247 262 MR0367131 10.1090/S0002-9947-1975-0367131-6 ZBL0307.26012