Research Article Existence of Solutions for Generalized Vector Quasi-Equilibrium

advertisement
Hindawi Publishing Corporation
Abstract and Applied Analysis
Volume 2013, Article ID 916089, 8 pages
http://dx.doi.org/10.1155/2013/916089
Research Article
Existence of Solutions for Generalized Vector Quasi-Equilibrium
Problems by Scalarization Method with Applications
De-ning Qu1,2 and Cao-zong Cheng1
1
2
College of Applied Science, Beijing University of Technology, Beijing 100124, China
College of Mathematics, Jilin Normal University, Siping, Jilin 136000, China
Correspondence should be addressed to De-ning Qu; qudening@126.com
Received 26 November 2012; Accepted 6 February 2013
Academic Editor: Ngai-Ching Wong
Copyright © 2013 D.-n. Qu and C.-z. Cheng. 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.
The aim of this paper is to study generalized vector quasi-equilibrium problems (GVQEPs) by scalarization method in locally convex
topological vector spaces. A general nonlinear scalarization function for set-valued mappings is introduced, its main properties are
established, and some results on the existence of solutions of the GVQEPs are shown by utilizing the introduced scalarization
function. Finally, a vector variational inclusion problem is discussed as an application of the results of GVQEPs.
1. Introduction
Recently, various vector equilibrium problems were investigated by adopting many different methods, such as the
scalarization method (e.g., [1, 2]), the recession method (e.g.,
[3]), and duality method (e.g., [4]).
The scalarization method is an important and efficacious tool of translating the vector problems into the scalar
problems. In 1992, Chen [5] translated a vector variational
inequality into a classical variational inequality by providing
a kind of solution conceptions with variable domination
structures. Gerth and Weidner [6] solved a vector optimization problem by introducing a scalarization function with a
variable and Gong [1] dealt with vector equilibrium problems by using the scalarization function defined in [6]. By
constructing new nonlinear scalarization functions with two
variables, Chen and Yang [7] discussed a vector variational
inequality and Chen et al. [2] investigated a generalized
vector quasi-equilibrium problem (GVQEP), respectively.
In addition, the authors in [8, 9] studied the systems of
vector equilibrium problems by the scalarization method
since the gap functions, indeed established by the nonlinear
scalarization function defined in [2], were adopted.
In this paper, we will discuss the GVQEPs by utilizing
scalarization method. The essential preliminaries are listed
in Section 2. On the basis of the works in [2, 6, 7], a general
nonlinear scalarization function of a set-valued mapping is
produced under a variable ordering structure and its main
properties are discussed in Section 3. The results of properties
for the general nonlinear scalarization function generalized
the corresponding ones in [2]. In Section 4, some results
on the existence of solutions of the GVQEPs are proved by
employing the scalarization function introduced in Section 3.
The GVQEPs are different from the one in [10] and include
the one in [2] as a special case. It is worth mentioning that
the existence results of solutions for the GVQEPs extend
the corresponding one in [2]. Finally, a vector variational
inclusion problem (VVIP) is given as an application of the
GVQEPs in Section 5.
Suppose that 𝑋 and π‘Œ are topological vector spaces
(TVSs). The subset 𝐷 ⊂ π‘Œ is called a cone, if πœ†π‘₯ ∈ 𝐷 for all
π‘₯ ∈ 𝐷 and πœ† > 0. A cone 𝐷 ⊂ π‘Œ is said to be proper, if 𝐷 =ΜΈ π‘Œ.
𝐴 is called 𝐷-closed [11] if 𝐴 + 𝐾 is closed and 𝐷-bonded [11]
if for each neighborhood π‘ˆ of zero in π‘Œ, there exists πœ† > 0
such that 𝐴 ⊂ π‘ˆ + 𝐷. A set-valued mapping 𝐺 : 𝑋 → 2π‘Œ is
called strict, if 𝐺(π‘₯) =ΜΈ 0 for any π‘₯ ∈ 𝑋. Throughout this paper,
R denotes by the set of the real numbers. Several notations
are also listed as follows:
𝐺 (𝐴) = ⋃ 𝐺 (𝑒) ,
𝑒∈𝐴
2
Abstract and Applied Analysis
π‘₯ − 𝐡 = {π‘₯ − 𝑦 : 𝑦 ∈ 𝐡} ,
𝐴 − 𝐡 = {π‘₯ − 𝑦 : π‘₯ ∈ 𝐴, 𝑦 ∈ 𝐡} ,
(1)
where 𝐴, 𝐡 ⊂ 𝑋 are nonempty, π‘₯ ∈ 𝑋 and 𝐺 : 𝑋 → 2π‘Œ .
Incidentally, every TVS is Hausdorff (see [12]).
2. Preliminaries
Let 𝑋 and π‘Œ be topological spaces and 𝐸 ⊂ 𝑋 a nonempty
subset. A function 𝑔 : 𝑋 → R is said to be upper
semicontinuous (usc for brevity) on 𝑋, if {𝑒 ∈ 𝑋 : 𝑔(𝑒) < πœ†}
is open for each πœ† ∈ R; to be lower semicontinuous (lsc for
brevity), {𝑒 ∈ 𝑋 : 𝑔(𝑒) > πœ†} is open for each πœ† ∈ R. In
addition, some known notions of continuity and closeness
for a set-valued mapping are given (see [13]). A set-valued
mapping 𝐺 : 𝐸 → 2π‘Œ is said to be usc at 𝑒0 ∈ 𝐸,
if for any neighborhood 𝑁(𝐺(𝑒0 )) of 𝐺(𝑒0 ), there exists a
neighborhood 𝐡(𝑒0 ) of 𝑒0 such that 𝐺(𝑒) ⊂ 𝑁(𝐺(𝑒0 )) for all
𝑒 ∈ 𝐡(𝑒0 ); to be lsc at 𝑒0 ∈ 𝐸, if for any 𝑦0 ∈ 𝐺(𝑒0 ) and any
neighborhood 𝑁(𝑦0 ) of 𝑦0 , there exists a neighborhood 𝐡(𝑒0 )
of 𝑒0 such that 𝐺(𝑒) ∩ 𝑁(𝑦0 ) =ΜΈ 0 for all 𝑒 ∈ 𝐡(𝑒0 ); to be usc
(resp., lsc) on 𝐸, if 𝐺 is usc (resp., lsc) at each 𝑒 ∈ 𝐸; to be
continuous at 𝑒0 ∈ 𝐸 (resp., on 𝐸), if 𝐺 is usc and lsc at 𝑒0
(resp., on 𝐸); to be closed, if its graph Graph(𝐺) = {(𝑒, 𝑦) ∈
𝐸 × π‘Œ : 𝑦 ∈ 𝐺(𝑒)} is closed in 𝐸 × π‘Œ.
Let π‘Œ and 𝑍 be real TVSs, 𝐴 ⊂ π‘Œ a nonempty subset, and
Μ‚ ∈ 𝐴 is called, vector minimal point
𝐷 a convex cone in π‘Œ. 𝑦
Μ‚ ∉ −𝐷 \ {0}
(resp., weakly vector minimal point) of 𝐴, if 𝑦 − 𝑦
Μ‚ ∉ − int 𝐷) for each 𝑦 ∈ 𝐴. The set of vector
(resp., 𝑦 − 𝑦
minimal points (resp., weakly vector minimal points) of 𝐴 is
denoted by Min𝐷𝐴 (resp., 𝑀Min𝐷𝐴).
Definition 1. Let 𝐡 ⊂ 𝑍 be nonempty convex subset. 𝐺 : 𝐡 →
2π‘Œ is called generalized 𝐷-quasiconvex, if for any 𝑦 ∈ π‘Œ, the
set {𝑒 ∈ 𝐡 : 𝐺(𝑒) ⊂ 𝑦 − 𝐷} is convex (here, 0 is regarded as a
convex set).
The generalized 𝐷-quasiconvexity and the 𝐷-quasiconvexity introduced in [14] for the set-valued mappings are both
generalizations of 𝐷-quasiconvexity for the single-valued
mapping introduced in [11], but they are indeed distinct. See
the following example.
Example 2. Let π‘Œ = R2 , 𝑍 = 𝐡 = R, and 𝐷 = R2+ and define
𝐺1 , 𝐺2 : 𝐡 → 2π‘Œ as
𝐺1 (𝑒) = {(π‘₯1 , π‘₯2 ) : π‘₯1 = 𝑒, −𝑒2 ≤ π‘₯2 ≤ 0} ,
𝐺2 (𝑒) = {(π‘₯1 , π‘₯2 ) : π‘₯1 = 𝑒, 0 ≤ π‘₯2 ≤ |sin 𝑒|} .
are convex. But the set
{𝑒 ∈ B : 𝐺1 (𝑒) ∩ ((2, −1) − 𝐷) =ΜΈ 0} = (−∞, −1] ∪ [1, 2]
(4)
is not convex and neither is the set
{𝑒 ∈ 𝐡 : 𝐺2 (𝑒) ⊂ (0, 0) − 𝐷} = {−π‘›πœ‹ : 𝑛 = 0, 1, 2, . . .} . (5)
Lemma 3 (see [15]). Let 𝑋 and π‘Œ be topological spaces and
𝑆 : 𝑋 → 2π‘Œ a set-valued mapping.
(1) If 𝑆 is usc with closed values, then 𝑆 is closed.
(2) If 𝑋 is compact and 𝑆 is usc with compact values, then
𝑆(𝑋) is compact.
Lemma 4 (see [13]). Let 𝑋 and π‘Œ be Hausdorff topological
spaces and 𝐸 ⊂ 𝑋 a nonempty compact set and let β„Ž : 𝐸 × π‘Œ →
R be a function and 𝐺 : 𝐸 → 2π‘Œ a set-valued mapping. If
β„Ž is continuous on 𝐸 × π‘Œ and 𝐺 is continuous with compact
values, then the marginal function 𝑉(𝑒) = sup𝑦∈𝐺(𝑒) β„Ž(𝑒, 𝑦) is
continuous and the marginal set-valued mapping 𝛿(𝑒) = {𝑦 ∈
𝐺(𝑒) : 𝑉(𝑒) = β„Ž(𝑒, 𝑦)} is usc.
Lemma 5 (Kakutani, see [13, Theorem 13 in Section 4 Chapter
6]). Let 𝐸 be a nonempty compact and convex subset of a
locally convex TVS 𝑋. If Φ : 𝐸 → 2𝐸 is usc and Φ(π‘₯) is a
nonempty, convex, and closed subset for any π‘₯ ∈ 𝐸, then there
exists π‘₯ ∈ 𝐸 such that π‘₯ ∈ Φ(π‘₯).
3. A General Nonlinear Scalarization Function
From now on, unless otherwise specified, let 𝑋, π‘Œ, and 𝑍 be
real TVSs and 𝐸 ⊂ 𝑋 and 𝐹 ⊂ 𝑍 nonempty subsets. Let 𝐢 :
𝐸 → 2π‘Œ be a set-valued mapping such that for any π‘₯ ∈ 𝐸,
𝐢(π‘₯) is a proper, closed, and convex cone with int 𝐢(π‘₯) =ΜΈ 0.
In this section, suppose that 𝐺 : 𝐹 → 2π‘Œ is a strict
mapping with compact values and 𝑒 : 𝐸 → π‘Œ is a
vector-valued mapping with 𝑒(π‘₯) ∈ int 𝐢(π‘₯), for all π‘₯ ∈
𝐸. Obviously, 𝐺(𝑒) is 𝐢(π‘₯)-closed [11, Definition 3.1 and
Proposition 3.3] for each π‘₯ ∈ 𝐸 and 𝑒 ∈ 𝐹. Then, in
view of [16, Lemma 3.1], we can define a general nonlinear
scalarization function of G as follows.
(2)
Then 𝐺1 is generalized 𝐷-quasiconvex but not 𝐷-quasiconvex, while 𝐺2 is just the reverse. Indeed, for any fixed 𝑦 =
(π‘₯1 , π‘₯2 ) ∈ π‘Œ, both the sets
0,
{
{
{ if π‘₯2 < 0, π‘₯1 ∈ R,
{𝑒 ∈ 𝐡 : 𝐺1 (𝑒) ⊂ 𝑦 − 𝐷} = {
{
{ (−∞, π‘₯1 ] ,
{ otherwise,
0,
{
{
{ if π‘₯2 < 0, π‘₯1 ∈ R,
{𝑒 ∈ 𝐡 : 𝐺2 (𝑒) ∩ (𝑦 − 𝐷) =ΜΈ 0} = {
{
{ (−∞, π‘₯1 ] ,
{ otherwise
(3)
Definition 6. The general nonlinear scalarization function
πœ‰πΊ : 𝐸 × πΉ → R, of 𝐺 is defined as
πœ‰πΊ (π‘₯, 𝑒) = min {πœ† ∈ R : 𝐺 (𝑒) ⊂ πœ†π‘’ (π‘₯) − 𝐢 (π‘₯)} ,
∀ (π‘₯, 𝑒) ∈ 𝐸 × πΉ.
(6)
Remark 7. If 𝑋 = π‘Œ = 𝑍 = 𝐸 = 𝐹 and 𝐺(𝑒) = {𝑒}, for all 𝑒 ∈
𝐹, then the general nonlinear scalarization function πœ‰πΊ of 𝐺
becomes the nonlinear scalarization function defined in [2].
Abstract and Applied Analysis
3
Example 8. Let 𝑋 = 𝑍 = R, π‘Œ = R2 , and 𝐸 = 𝐹 = [0, +∞) ⊂
𝑋 and let 𝐺, 𝐢 : 𝐸 → 2π‘Œ , 𝑒 : 𝐸 → π‘Œ define as
𝐺 (𝑒) = {(π‘ˆ, 𝑉) : π‘ˆ − 𝑒 ≤ 𝑉 ≤ 𝑒 − π‘ˆ, 0 ≤ π‘ˆ ≤ 𝑒} ,
3
𝐢 (π‘₯) = {(π‘ˆ, 𝑉) : 0 ≤ 𝑉 ≤ ( + π‘₯) π‘ˆ, π‘ˆ ≥ 0} ,
2
𝑒 (π‘₯) = (1, 1) ,
∀𝑒 ∈ 𝐹,
∀π‘₯ ∈ 𝐸,
∀π‘₯ ∈ 𝐸,
(7)
respectively. We can see that πœ‰πΊ(π‘₯, 𝑒) = (1 + (2/(1 + 2π‘₯)))𝑒.
Proof. Letting 𝑒1 , 𝑒2 ∈ 𝐹 such that 𝐺(𝑒1 ) − 𝐺(𝑒2 ) ⊂ int 𝐢(π‘₯)
and πœ† = πœ‰πΊ(π‘₯, 𝑒1 ), we have
𝐺 (𝑒2 ) ⊂ 𝐺 (𝑒1 ) − int 𝐢 (π‘₯)
⊂ πœ†π‘’ (π‘₯) − 𝐢 (π‘₯) − int 𝐢 (π‘₯) (by Definition 6) (9)
⊂ πœ†π‘’ (π‘₯) − int 𝐢 (π‘₯) .
It is from this assertion that πœ‰πΊ(π‘₯, 𝑒2 ) < πœ† = πœ‰πΊ(π‘₯, 𝑒1 ) by
Theorem 12 (1).
Theorem 14. Suppose that 𝑒 is continuous.
Definition 9. Let 𝐷 ⊂ π‘Œ be a cone, 𝐻 : 𝐹 → 2π‘Œ , and 𝜍 : 𝐹 →
R. 𝜍 is called monotone (resp., strictly monotone) with respect
to (wrt for brevity) 𝐻, if for any 𝑒1 , 𝑒2 ∈ 𝐹, 𝐻(𝑒1 ) − 𝐻(𝑒2 ) ⊂
int 𝐷 implies that 𝜍(𝑒2 ) ≤ 𝜍(𝑒1 ) (resp., 𝜍(𝑒2 ) < 𝜍(𝑒1 )).
(1) If π‘Š and 𝐺 are usc on 𝐸 and 𝐹, respectively, where
Obviously, the strict monotonicity wrt 𝐻 implies the
monotonicity wrt 𝐻. Moreover, if 𝐻(𝑒) = {𝑒}, for all 𝑒 ∈ 𝐹,
then the (strict) monotonicity wrt 𝐻 of 𝜍 is equivalent to
the (strict) monotonicity under the general order structures.
The following examples illuminate the relationship between
the monotonicity wrt 𝐻 and the monotonicity in the normal
sense when 𝐻 is not the identity mapping.
(2) If 𝐢 is usc on 𝐸 and 𝐺 is lsc on 𝐹, then πœ‰πΊ is lsc on 𝐸 × πΉ.
Example 10. Let π‘Œ = 𝑍 = R, 𝐹 = [0, 10], and 𝐷 = R+ and let
𝑒, 𝑒 ∈ [0, 5] ,
𝜍 (𝑒) = {
0, 𝑒 ∈ (5, 10] ,
[𝑒, 𝑒 + 1] , 𝑒 ∈ [0, 5) ,
{
{
𝐻 (𝑒) = {[0, 6] ,
𝑒 = 5,
{
10
−
𝑒]
,
𝑒
∈ (5, 10] .
[0,
{
(8)
Then 𝐻(𝑒1 ) − 𝐻(𝑒2 ) ⊂ int 𝐷 implies that 𝜍(𝑒1 ) > 𝜍(𝑒2 ). Thus
𝜍 is strictly monotone wrt 𝐻, while 𝜍 is not monotone in the
normal sense.
Example 11. Let π‘Œ = 𝑍 = R, 𝐹 = [0, 5], and 𝐷 = R+ and let
𝜍(𝑒) = 𝑒, for all 𝑒 ∈ 𝐹, and 𝐻(𝑒) = [−𝑒−1, −𝑒], for all 𝑒 ∈ 𝐹.
Then 𝜍 is strictly monotone in the normal sense, but 𝜍 is not
monotone wrt 𝐻. As the case stands, 𝐻(𝑒1 ) − 𝐻(𝑒2 ) ⊂ int 𝐷
is equivalent to 𝑒1 , 𝑒2 ∈ 𝐹 and 𝑒2 − 𝑒1 > 1, which just results
in 𝜍(𝑒1 ) < 𝜍(𝑒2 ).
Now some main properties of the general nonlinear
scalarization function are established. First, according to [16,
Proposition 3.1], we have the following.
Theorem 12. For each πœ† ∈ R, π‘₯ ∈ 𝐸 and 𝑒 ∈ 𝐹, the following
assertions hold.
(1) πœ‰πΊ(π‘₯, 𝑒) < πœ† ⇔ 𝐺(𝑒) ⊂ πœ†π‘’(π‘₯) − int 𝐢(π‘₯).
(2) πœ‰πΊ(π‘₯, 𝑒) ≤ πœ† ⇔ 𝐺(𝑒) ⊂ πœ†π‘’(π‘₯) − 𝐢(π‘₯).
Theorem 13. πœ‰πΊ is strictly monotone wrt 𝐺 in the second
variable.
π‘Š (π‘₯) = π‘Œ \ int 𝐢 (π‘₯) ,
∀π‘₯ ∈ 𝐸,
(10)
then πœ‰πΊ is usc on 𝐸 × πΉ.
Proof. (1) It is sufficient to attest the fact that for each πœ† ∈ R,
the set
𝐴 = {(π‘₯, 𝑒) ∈ 𝐸 × πΉ : πœ‰πΊ (π‘₯, 𝑒) ≥ πœ†}
(11)
is closed. As a matter of fact, for any sequence {(π‘₯π‘˜ , π‘’π‘˜ )} in
𝐴 such that (π‘₯π‘˜ , π‘’π‘˜ ) → (π‘₯0 , 𝑒0 ) as π‘˜ → ∞, πœ‰πΊ(π‘₯π‘˜ , π‘’π‘˜ ) ≥
πœ†, that is, 𝐺(π‘’π‘˜ ) ⊂ΜΈ πœ†π‘’(π‘₯π‘˜ ) − int 𝐢(π‘₯π‘˜ ), by Theorem 12 (1).
Obviously, 𝐹0 = {𝑒, 𝑒1 , 𝑒2 , . . .} ⊂ 𝐹 is compact and so is 𝐺(𝐹0 )
by Lemma 3 (2). Then there exists π‘¦π‘˜ ∈ 𝐺(π‘’π‘˜ ) ⊂ 𝐺(𝐹0 ) such
that π‘¦π‘˜ ∉ πœ†π‘’(π‘₯π‘˜ ) − int 𝐢(π‘₯π‘˜ ) and a subsequence {π‘¦π‘˜π‘– } of {π‘¦π‘˜ },
such that π‘¦π‘˜π‘– → 𝑦0 ∈ 𝐺(𝐹0 ) as 𝑖 → ∞ and
πœ†π‘’ (π‘₯π‘˜π‘– ) − π‘¦π‘˜π‘– ∉ int 𝐢 (π‘₯π‘˜π‘– ) ,
that is, πœ†π‘’ (π‘₯π‘˜π‘– ) − π‘¦π‘˜π‘– ∈ π‘Š (π‘₯π‘˜π‘– ) .
(12)
Obviously, 𝐺 is usc with closed values, which implies that 𝐺
is closed by Lemma 3 (1). Hence, 𝑦0 ∈ 𝐺(𝑒0 ). Similarly, π‘Š is
closed. Letting 𝑖 → ∞ in (12) and applying the continuity
of 𝑒 and the closeness of π‘Š, we obtain that πœ†π‘’(π‘₯0 ) − 𝑦0 ∈
π‘Š(π‘₯0 ). Namely, 𝑦0 ∉ πœ†π‘’(π‘₯0 ) − int 𝐢(π‘₯0 ). Thus 𝐺(𝑒0 ) ⊂ΜΈ
πœ†π‘’(π‘₯0 ) − int 𝐢(π‘₯0 ), which is equivalent to πœ‰πΊ(π‘₯0 , 𝑒0 ) ≥ πœ† by
Theorem 12 (1). Therefore, πœ† ∈ 𝐴 and 𝐴 is closed.
(2) It’s enough to argue that for each πœ† ∈ R, the set
𝐡 = {(π‘₯, 𝑒) ∈ 𝐸 × πΉ : πœ‰πΊ (π‘₯, 𝑒) ≤ πœ†}
(13)
is closed. In fact, for any sequence {(π‘₯π‘˜ , π‘’π‘˜ )} in 𝐡 such
that (π‘₯π‘˜ , π‘’π‘˜ ) → (π‘₯0 , 𝑒0 ) as π‘˜ → ∞, by Theorem 12 (2),
πœ‰πΊ(π‘₯π‘˜ , π‘’π‘˜ ) ≤ πœ†, equivalently, 𝐺(π‘’π‘˜ ) ⊂ πœ†π‘’(π‘₯π‘˜ ) − 𝐢(π‘₯π‘˜ ). For
any 𝑦0 ∈ 𝐺(𝑒0 ), there exists π‘¦π‘˜ ∈ 𝐺(π‘’π‘˜ ) such that π‘¦π‘˜ → 𝑦0
as π‘˜ → ∞ according to the equivalent definition of lower
semicontinuity of 𝐺 (see [13], page 108). Also,
πœ†π‘’ (π‘₯π‘˜ ) − π‘¦π‘˜ ∈ 𝐢 (π‘₯π‘˜ ) .
(14)
Linking the continuity of 𝑒 with the closeness of 𝐢 inferred
from Lemma 3 (1) and letting π‘˜ → ∞ in (14), we get πœ†π‘’(π‘₯0 )−
𝑦0 ∈ 𝐢(π‘₯0 ) and 𝐺(𝑒0 ) ⊂ πœ†π‘’(π‘₯0 ) − 𝐢(π‘₯0 ). Consequently,
4
Abstract and Applied Analysis
Theorem 12 (2) leads to πœ‰πΊ(π‘₯0 , 𝑒0 ) ≤ πœ† which amounts to
(π‘₯0 , 𝑒0 ) ∈ 𝐡. 𝐡 is closed.
If 𝑋 = π‘Œ = 𝑍 = 𝐸 = 𝐹 (𝐹 is not required to be
compact) and 𝐺(𝑒) = {𝑒}, for all 𝑒 ∈ 𝐹, then Theorem 14
becomes [2, Theorem 2.1]. Actually, [2, Theorem 2.1] can be
regarded as the case where 𝐺 is continuous in Theorem 14.
In addition, Example 2.1 (resp., Example 2.2) in [2] shows
that if 𝐢 (resp., π‘Š) is not usc, maybe the general nonlinear
scalarization function fails to be lsc (resp., usc) under all the
other assumptions. Now the following example demonstrates
that the assumption of the upper semicontinuity (resp., lower
semicontinuity) of 𝐺 is necessary in Theorem 14 (1) (resp., (2))
even if 𝐢 is continuous.
Example 15. Let 𝑋 = 𝑍 = R, Y = R2 , 𝐸 = R+ , and 𝐹 =
[−10, 10] ⊂ 𝑍, and let
𝑒 (π‘₯) = (1, 1) ,
∀π‘₯ ∈ 𝐸,
3
𝐢 (π‘₯) = {(π‘ˆ, 𝑉) : 0 ≤ 𝑉 ≤ ( + π‘₯) π‘ˆ, π‘ˆ ≥ 0} ,
2
(15)
∀π‘₯ ∈ 𝐸.
(16)
GVQEP2: Find π‘₯ ∈ 𝑃(π‘₯) and 𝑧 ∈ 𝑄(π‘₯) such that
𝑓 (V, π‘₯, 𝑧) − 𝑓 (π‘₯, π‘₯, 𝑧)
⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
∀V ∈ 𝑃 (π‘₯) ,
where 𝑃 : 𝐸 → 2𝐸 , 𝑄 : 𝐸 → 2𝐹 , and 𝑓 : 𝐸×𝐸×𝐹 →
2π‘Œ are set-valued mappings.
It’s worth noting that the GVQEP considered in [2] is just
the special case of the GVQEP2 (when 𝑓 is single valued).
Lemma 16. Let 𝑒 : 𝐸 → π‘Œ be a vector-valued mapping such
that 𝑒(π‘₯) ∈ int 𝐢(π‘₯) for all π‘₯ ∈ 𝐸 and 𝑓 : 𝐸 × πΉ → 2π‘Œ
a set-valued mapping and define 𝑓𝑧 (π‘₯) = 𝑓(π‘₯, 𝑧) for each
𝑧 ∈ 𝐹. If for each 𝑧 ∈ 𝐹, 𝑓𝑧 is generalized 𝐢(𝑔(𝑧))-quasiconvex
with compact values, then π‘₯ 󳨃→ β„Ž(π‘₯, 𝑧) is R+ -quasiconvex,
where β„Ž(π‘₯, 𝑧) = πœ‰π‘“π‘§ (𝑔(𝑧), π‘₯) and πœ‰π‘“π‘§ is the general nonlinear
scalarization function of 𝑓𝑧 .
Proof. It’s sufficient to testify that 𝐿 𝑧 (πœ†) ⊂ 𝑋 is convex, where
𝐿 𝑧 (πœ†) = {π‘₯ ∈ 𝐸 : πœ‰π‘“π‘§ (𝑔 (𝑧) , π‘₯) ≤ πœ†}
(1) Define
𝐺 (𝑒) = {
[𝑒, 𝑒 + 1] , 𝑒 ∈ (0, 10] ,
0,
𝑒 ∈ [−10, 0] .
(17)
𝑒 + 1,
πœ‰πΊ (π‘₯, 𝑒) = {
0,
π‘₯ ∈ 𝐸, 0 < 𝑒 ≤ 10,
π‘₯ ∈ 𝐸, −10 ≤ 𝑒 ≤ 0.
πœ‰πΊ is not usc on 𝐸 × πΉ due to the fact that {(π‘₯, 𝑒) ∈ 𝐸 × πΉ :
πœ‰(π‘₯, 𝑒) ≥ 1} = R+ × (0, 10] is not closed.
(2) Consider the following mapping:
[𝑒, 𝑒 + 1] , 𝑒 ∈ [0, 10] ,
𝐺 (𝑒) = {
0,
𝑒 ∈ [−10, 0) .
(19)
Obviously, 𝐺 is not lsc on 𝐹. Also, πœ‰πΊ fails to be lsc on 𝐸 × πΉ,
where
𝑒 + 1,
πœ‰πΊ (π‘₯, 𝑒) = {
0,
π‘₯ ∈ 𝐸, 0 ≤ 𝑒 ≤ 10,
π‘₯ ∈ 𝐸, −10 ≤ 𝑒 < 0.
(20)
4. Existence Results on Solutions of
the GVQEPs
In this section, further suppose that 𝑋, π‘Œ, and 𝑍 are locally
convex. Let 𝑔 : 𝐹 → 𝐸 be a vector-valued mapping. Now
two GVQEPs are characterized as follows:
GVQEP1: Seek π‘₯ ∈ 𝑃(π‘₯) and 𝑧 ∈ 𝑄(π‘₯) such that
𝑓 (V, 𝑧) − 𝑓 (π‘₯, 𝑧)
⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
∀V ∈ 𝑃 (π‘₯) ,
𝑖 = 1, 2,
that is, 𝑓𝑧 (π‘₯𝑖 ) ⊂ πœ†π‘’ (𝑔 (𝑧)) − 𝐢 (𝑔 (𝑧)) ,
(18)
(23)
for each 𝑧 ∈ 𝐹 and πœ† ∈ R. Indeed, for any π‘₯1 , π‘₯2 ∈ 𝐿 𝑧 (πœ†),
πœ‰π‘“π‘§ (𝑔 (𝑧) , π‘₯𝑖 ) ≤ πœ†,
Evidently, 𝐺 is not usc on 𝐹. After simply calculating,
(22)
𝑖 = 1, 2.
(24)
Clearly, for each 𝑧 ∈ 𝐸, π‘₯1 , π‘₯2 ∈ 𝑀, where
𝑀 = {π‘₯ ∈ 𝐸 : 𝑓𝑧 (π‘₯) ⊂ πœ†π‘’ (𝑔 (𝑧)) − 𝐢 (𝑔 (𝑧))} .
(25)
𝑀 is convex by reason of the generalized 𝐢(𝑔(𝑧))Μƒ = 𝑑π‘₯1 + (1 − 𝑑)π‘₯2 ∈ 𝑀
quasiconvexity of 𝑓𝑧 and so π‘₯
for all 0 ≤ 𝑑 ≤ 1. Hence,
𝑓𝑧 (Μƒ
π‘₯) ⊂ πœ†π‘’ (𝑔 (𝑧)) − 𝐢 (𝑔 (𝑧)) ,
(26)
Μƒ ) ≤ πœ† by Theorem 12 (2). Thus,
which is equal to πœ‰π‘“π‘§ (𝑔(𝑧), π‘₯
Μƒ ∈ 𝐿 𝑧 (πœ†) and 𝐿 𝑧 (πœ†) is convex.
π‘₯
Now a result on existence of solutions of the GVQEP1 is
verified by making use of the general nonlinear scalarization
function defined in Section 3.
Theorem 17. Let 𝐸 and 𝐹 be compact and convex subsets and
𝑃 : 𝐸 → 2𝐸 , 𝑄 : 𝐸 → 2𝐹 , and 𝑓 : 𝐸 × πΉ → 2π‘Œ set-valued
mappings. For each 𝑧 ∈ 𝐹, define 𝑓𝑧 (π‘₯) = 𝑓(π‘₯, 𝑧). Suppose that
the following conditions are fulfilled:
(a) π‘₯ 󳨃→ int 𝐢(π‘₯) has a continuous select 𝑒 : 𝐸 → π‘Œ;
(b) both 𝐢 and π‘Š are usc on 𝐸, where
(21)
where 𝑃 : 𝐸 → 2𝐸 , 𝑄 : 𝐸 → 2𝐹 , and 𝑓 : 𝐸 × πΉ →
2π‘Œ are set-valued mappings.
π‘Š (π‘₯) = π‘Œ \ int 𝐢 (π‘₯) ,
∀π‘₯ ∈ 𝐸;
(27)
(c) 𝑔 is continuous, 𝑓 and 𝑃 are strict and continuous, and
𝑄 is strict and usc;
(d) for each 𝑧 ∈ 𝐹, 𝑓𝑧 is generalized 𝐢(𝑔(𝑧))-quasiconvex;
Abstract and Applied Analysis
5
(e) for each (π‘₯, 𝑧) ∈ 𝐸 × πΉ, 𝑓(π‘₯, 𝑧) is compact and for each
π‘₯ ∈ 𝐸, both 𝑃(π‘₯) and 𝑄(π‘₯) are closed and convex.
Then the GVQEP1 has a solution (π‘₯, 𝑧) ∈ 𝐸 × πΉ; that is, there
exist π‘₯ ∈ 𝑃(π‘₯) and 𝑧 ∈ 𝑄(π‘₯) such that
𝑓 (V, 𝑧) − 𝑓 (π‘₯, 𝑧) ⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
∀V ∈ 𝑃 (π‘₯) .
(28)
Further assume that 𝑓(π‘₯, 𝑧) ⊂ 𝐢(𝑔(𝑧)) for each π‘₯ ∈ 𝐸 and
𝑧 ∈ 𝑄(π‘₯). Then
𝑓 (V, 𝑧) ⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
∀V ∈ 𝑃 (π‘₯) .
(29)
Proof. Denote 𝐾 = 𝐸 × πΉ and 𝑀 = (π‘₯, 𝑧) where π‘₯ ∈ 𝐸 and
Μƒ : 𝐾 → 𝐸 as
𝑧 ∈ 𝐹, and define 𝑔
Μƒ (𝑀) = 𝑔
Μƒ (π‘₯, 𝑧) = 𝑔 (𝑧) .
𝑔
(30)
Μƒ is continuous on 𝐾. The general nonlinear
Obviously, 𝑔
scalarization function of 𝑓
πœ‰π‘“ (𝑒, 𝑀) = min {πœ† ∈ R : 𝑓 (𝑀) ⊂ πœ†π‘’ (𝑒) − 𝐢 (𝑒)}
(31)
is continuous on 𝐸 × πΎ by Theorem 14. Set β„Ž(π‘₯, 𝑧) =
πœ‰π‘“π‘§ (𝑔(𝑧), π‘₯), where πœ‰π‘“π‘§ is the general nonlinear scalarization
function of 𝑓𝑧 . Since
β„Ž (𝑀) = β„Ž (π‘₯, 𝑧)
= min {πœ† ∈ R : 𝑓𝑧 (π‘₯) ⊂ πœ†π‘’ (𝑔 (𝑧)) − 𝐢 (𝑔 (𝑧))}
= min {πœ† ∈ R : 𝑓 (π‘₯, 𝑧) ⊂ πœ†π‘’ (Μƒ
𝑔 (π‘₯, 𝑧)) − 𝐢 (Μƒ
𝑔 (π‘₯, 𝑧))}
𝑔 (𝑀) , 𝑀) ,
= πœ‰π‘“ (Μƒ
(32)
Μƒ.
β„Ž is continuous on 𝐾 by virtue of the continuity of πœ‰π‘“ and 𝑔
Μƒ : 𝐸 × πΉ → 2𝐸×𝐹 as
Define 𝑃
∀ (π‘₯, 𝑧) ∈ 𝐸 × πΉ.
πœ† = min β„Ž (V, 𝑧) .
Then for any 𝑒1 , 𝑒2 ∈ πœ‚(π‘₯, 𝑧), β„Ž(𝑒𝑖 , 𝑧) = πœ†, 𝑖 = 1, 2. Clearly,
Μƒ = 𝑑𝑒1 + (1 − 𝑑)𝑒2 ∈ 𝑃(π‘₯) for all 𝑑 ∈ [0, 1] since 𝑃(π‘₯) is
𝑒
convex. In view of condition (d) and Lemma 16, π‘₯ 󳨃→ β„Ž(π‘₯, 𝑧)
is R+ -quasiconvex, which deduces that for each 𝑧 ∈ 𝐹, the set
𝐿 𝑧 (πœ†) = {π‘₯ ∈ 𝐸 : β„Ž (π‘₯, 𝑧) ≤ πœ†}
(ii) For each (π‘₯, 𝑧) ∈ 𝐸 × πΉ and for any sequence {π‘’π‘˜ } ⊂
πœ‚(π‘₯, 𝑧) such that π‘’π‘˜ → 𝑒0 as π‘˜ → ∞,
β„Ž (π‘’π‘˜ , 𝑧) = min β„Ž (V, 𝑧) ,
(38)
Since β„Ž is continuous,
β„Ž (𝑒0 , 𝑧) = lim β„Ž (π‘’π‘˜ , 𝑧) = min β„Ž (V, 𝑧) .
π‘˜→∞
V∈𝑃(π‘₯)
(39)
Thus 𝑒0 ∈ πœ‚(π‘₯, 𝑧) and πœ‚(π‘₯, 𝑧) is closed.
Now consider a set-valued mapping Φ : 𝐸 × πΉ → 2𝐸×𝐹
prescribed as
Φ (π‘₯, 𝑧) = πœ‚ (π‘₯, 𝑧) × π‘„ (π‘₯) ,
π‘₯ ∈ 𝑃 (π‘₯) ,
∀ (π‘₯, 𝑧) ∈ 𝐸 × πΉ.
(40)
𝑧 ∈ 𝑄 (π‘₯) ,
β„Ž (π‘₯, 𝑧) = min β„Ž (V, 𝑧) .
V∈𝑃(π‘₯)
πœ‰π‘“π‘§ (𝑔 (𝑧) , V) ≥ πœ‰π‘“π‘§ (𝑔 (𝑧) , π‘₯) ,
V∈𝑃(π‘₯)
− β„Ž (V, 𝑏)}.
𝑓 (V, 𝑧) − 𝑓 (π‘₯, 𝑧) = 𝑓𝑧 (V) − 𝑓𝑧 (π‘₯)
∀V ∈ 𝑃 (π‘₯) ,
(43)
by Theorem 13. The first conclusion is proved. Further assume
that 𝑓(π‘₯, 𝑧) ⊂ 𝐢(𝑔(𝑧)) for each π‘₯ ∈ 𝐸 and 𝑧 ∈ 𝑄(π‘₯). If the
second conclusion is not true; namely, there exists V ∈ 𝑃(π‘₯)
such that 𝑓(V, 𝑧) ⊂ − int 𝐢(𝑔(𝑧)), then
𝑓 (V, 𝑧) − 𝑓 (π‘₯, 𝑧) ⊂ − int 𝐢 (𝑔 (𝑧)) − 𝐢 (𝑔 (𝑧))
(34)
By Lemma 4, 𝛿(π‘₯, 𝑧) is usc. Let
(42)
which deduces that π‘₯ ∈ 𝑃(π‘₯), 𝑧 ∈ 𝑄(π‘₯), and
⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
= {(𝑒, 𝑧) ∈ 𝑃 (π‘₯) × {𝑧} : −β„Ž (𝑒, 𝑧) = max − β„Ž (V, 𝑧)}
(41)
Hence, for each V ∈ 𝑃(π‘₯), β„Ž(V, 𝑧) ≥ β„Ž(π‘₯, 𝑧). In other words,
V∈𝑃(π‘₯)
max
∀π‘˜.
V∈𝑃(π‘₯)
𝛿 (π‘₯, 𝑧) = {(𝑒, 𝑧) ∈ 𝑃 (π‘₯) × {𝑧} : β„Ž (𝑒, 𝑧) = min β„Ž (V, 𝑧)}
Μƒ
(V,𝑏)∈𝑃(π‘₯,𝑧)
(37)
is convex. Thus, β„Ž(Μƒ
𝑒, 𝑧) ≤ πœ†. It follows from the definition of πœ†
Μƒ ∈ πœ‚(π‘₯, 𝑧) and the convexity
that β„Ž(Μƒ
𝑒, 𝑧) = πœ†, which results in 𝑒
of πœ‚(π‘₯, 𝑧).
(33)
Μƒ has compact values and is continuous. Define a
Obviously, 𝑃
set-valued mapping 𝛿 : 𝐸 × πΉ → 2𝐸×𝐹 as
Μƒ (π‘₯, 𝑧) : −β„Ž (𝑒, π‘Ž) =
= {(𝑒, π‘Ž) ∈ 𝑃
(36)
V∈𝑃(π‘₯)
It’s easy to check that Φ is usc on 𝐸 × πΉ and for each (π‘₯, 𝑧) ∈
𝐸 × πΉ, Φ(π‘₯, 𝑧) is convex and closed. By Lemma 5, Φ exists a
fixed point (π‘₯, 𝑧), that is to say,
= min {πœ† ∈ R : 𝑓 (𝑀) ⊂ πœ†π‘’ (Μƒ
𝑔 (𝑀)) − 𝐢 (Μƒ
𝑔 (𝑀))}
Μƒ (π‘₯, 𝑧) = 𝑃 (π‘₯) × {𝑧} ,
𝑃
(i) Define
= − int 𝐢 (𝑔 (𝑧)) .
(44)
This is absurd.
πœ‚ (π‘₯, 𝑧) = {𝑒 ∈ 𝑃 (π‘₯) : β„Ž (𝑒, 𝑧) = min β„Ž (V, 𝑧)} .
V∈𝑃(π‘₯)
(35)
Then 𝛿(π‘₯, 𝑧) = πœ‚(π‘₯, 𝑧) × {𝑧}. The upper semicontinuity of πœ‚ is
obvious. Now we show that for each (π‘₯, 𝑧) ∈ 𝐸 × πΉ, πœ‚(π‘₯, 𝑧) is
convex and closed. Detailedly,
If mapping 𝐢 in Theorem 17 satisfies that for each 𝑦 ∈ π‘Œ,
(int 𝐢)−1 (𝑦) = {π‘₯ ∈ 𝐸 : 𝑦 ∈ int 𝐢(π‘₯)} is open, then
π‘₯ 󳨃→ int 𝐢(π‘₯) must exist a continuous selection by Browder
Selection Theorem [17]. Especially, when 𝑓 is single valued in
Theorem 17, a result is stated as follows.
6
Abstract and Applied Analysis
Corollary 18. Let 𝐸 and 𝐹 be compact and convex subsets, 𝑃 :
𝐸 → 2𝐸 and 𝑄 : 𝐸 → 2𝐹 set-valued mappings and 𝑓 :
𝐸 × πΉ → π‘Œ a vector-valued mapping. For each 𝑧 ∈ 𝐹, define
𝑓𝑧 (π‘₯) = 𝑓(π‘₯, 𝑧). Suppose that the following conditions are in
force:
(a) π‘₯ 󳨃→ int 𝐢(π‘₯) has a continuous select 𝑒 : 𝐸 → π‘Œ;
(b) both C and π‘Š are usc on 𝐸, where
π‘Š (π‘₯) = π‘Œ \ int 𝐢 (π‘₯) ,
∀π‘₯ ∈ 𝐸;
(45)
(c) 𝑓 and 𝑔 are continuous, 𝑃 is strict and continuous, and
𝑄 is strict and usc;
(d) for each 𝑧 ∈ 𝐹, 𝑓𝑧 is 𝐢(𝑔(𝑧))-quasiconvex;
(e) for each π‘₯ ∈ 𝐸, both 𝑃(π‘₯) and 𝑄(π‘₯) are closed and
convex.
Then there exist π‘₯ ∈ 𝑃(π‘₯) and 𝑧 ∈ 𝑄(π‘₯) such that
𝑓 (π‘₯, 𝑧) ∈ 𝑀Min𝐢(𝑔(𝑧)) 𝑓 (𝑃 (π‘₯) , 𝑧) .
(46)
Further suppose that 𝑓(π‘₯, 𝑧) ∈ 𝐢(𝑔(𝑧)) for each π‘₯ ∈ 𝐸 and
𝑧 ∈ 𝑄(π‘₯). Then
𝑓 (𝑃 (π‘₯) , 𝑧) ∩ (− int 𝐢 (𝑔 (𝑧))) = 0.
(47)
Theorem 19. Let 𝐸 and 𝐹 be compact and convex subsets and
𝑃 : 𝐸 → 2𝐸 , 𝑄 : 𝐸 → 2𝐹 , and 𝑓 : 𝐸×𝐸×𝐹 → 2π‘Œ set-valued
mappings. For each (π‘₯, 𝑧) ∈ 𝐸 × πΉ, define 𝑓π‘₯𝑧 (𝑒) = 𝑓(𝑒, π‘₯, 𝑧).
Suppose that the following conditions hold:
(a) π‘₯ 󳨃→ int 𝐢(π‘₯) has a continuous select 𝑒 : 𝐸 → π‘Œ;
(b) both 𝐢 and π‘Š are usc on 𝐸, where
π‘Š (π‘₯) = π‘Œ \ int 𝐢 (π‘₯) ,
∀π‘₯ ∈ 𝐸;
(48)
(c) 𝑔 is continuous, 𝑓 and 𝑃 are strict and continuous, and
𝑄 is strict and usc;
(d) for each (π‘₯, 𝑧) ∈ 𝐸 × πΉ, 𝑓π‘₯𝑧 is generalized 𝐢(𝑔(𝑧))quasiconvex;
(e) for each (𝑒, π‘₯, 𝑧) ∈ 𝐸 × πΈ × πΉ, 𝑓(𝑒, π‘₯, 𝑧) is compact
and for each π‘₯ ∈ 𝐸, both 𝑃(π‘₯) and 𝑄(π‘₯) are closed and
convex.
Then the GVQEP2 exists a solution (π‘₯, 𝑧) ∈ 𝐸 × πΉ; namely,
there exist π‘₯ ∈ 𝑃(π‘₯) and 𝑧 ∈ 𝑄(π‘₯) such that
𝑓 (V, π‘₯, 𝑧) − 𝑓 (π‘₯, π‘₯, 𝑧) ⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
∀V ∈ 𝑃 (π‘₯) .
(49)
The result below follows from Theorem 19 immediately by
further assuming that 𝑓 is single-valued.
Corollary 20 (see [2]). Let 𝐸 and 𝐹 be compact and convex
subsets, 𝑃 : 𝐸 → 2𝐸 and 𝑄 : 𝐸 → 2𝐹 set-valued mappings,
and 𝑓 : 𝐸 × πΈ × πΉ → π‘Œ a vector-valued mapping. For each
(π‘₯, 𝑧) ∈ 𝐸 × πΉ, define 𝑓π‘₯𝑧 (𝑒) = 𝑓(𝑒, π‘₯, 𝑧). The following
assumptions are in operation:
(a) π‘₯ 󳨃→ int 𝐢(π‘₯) has a continuous select 𝑒 : 𝐸 → π‘Œ;
(b) both 𝐢 and π‘Š are usc on 𝐸, where
π‘Š (π‘₯) = π‘Œ \ int 𝐢 (π‘₯) ,
∀V ∈ 𝑃 (π‘₯) .
(50)
Proof. Denoting 𝐾 = 𝐸 × πΉ, we see that 𝐾 is compact.
Μƒ : 𝐾 → 𝐸 and 𝑆 : 𝐸 → 2𝐾 as 𝑔
Μƒ (𝑀) =
Define 𝑔
Μƒ (π‘₯, 𝑧) = 𝑔(𝑧), for all 𝑀 = (π‘₯, 𝑧) ∈ 𝐾 and 𝑆(π‘₯) = {π‘₯} ×
𝑔
Μƒ is continuous on 𝐾
𝑄(π‘₯), for all π‘₯ ∈ 𝐸, respectively. Then 𝑔
𝑔(𝑀))-quasiconvex
and for each 𝑀 ∈ 𝐾, 𝑓𝑀 is generalized 𝐢(Μƒ
according to condition (d). Since 𝑄 is strict and usc with
compact values, so is 𝑆. Replacing 𝐹, 𝑔, and 𝑄 in Theorem 17
Μƒ , and 𝑆, respectively, we see that these conclusions are
by 𝐾, 𝑔
true.
(51)
(c) 𝑓 and 𝑔 are continuous, 𝑃 is strict and continuous, and
𝑄 is strict and usc;
(d) for each (π‘₯, 𝑧) ∈ 𝐸 × πΉ, 𝑓π‘₯𝑧 is 𝐢(𝑔(𝑧))-quasiconvex;
(e) for each π‘₯ ∈ 𝐸, both 𝑃(π‘₯) and 𝑄(π‘₯) are closed and
convex.
Then there exist π‘₯ ∈ 𝑃(π‘₯) and 𝑧 ∈ 𝑄(π‘₯) such that
𝑓 (π‘₯, π‘₯, 𝑧) ∈ 𝑀Min𝐢(𝑔(𝑧)) 𝑓 (𝑃 (π‘₯) , π‘₯, 𝑧) .
(52)
Additionally, on the assumption that 𝑓(π‘₯, π‘₯, 𝑧) ∈ 𝐢(𝑔(𝑧)) for
each π‘₯ ∈ 𝐸 and 𝑧 ∈ 𝑄(π‘₯),
𝑓 (𝑃 (π‘₯) , π‘₯, 𝑧) ∩ (− int 𝐢 (𝑔 (𝑧))) = 0.
(53)
5. An Application of GVQEP1: A VVIP
Let 𝑋 be a real 𝐹-space (𝑋 is called an 𝐹-space [14], if it is a
TVS such that its topology is induced by a complete invariant
metric), π‘Œ a real locally convex TVS, 𝐸 ⊂ 𝑋 a nonempty
subset and 𝐷 ⊂ π‘Œ a proper, closed, and convex cone with
int 𝐷 =ΜΈ 0 and let 𝑙 : 𝐸 → 2𝐿(𝑋,π‘Œ) be strict and continuous
with compact values. A VVIP is described as follows: to find
π‘₯ ∈ 𝐸 such that
βŸ¨π‘™ (π‘₯) , π‘₯ − π‘₯⟩ ⊂ΜΈ − int 𝐷,
Furthermore, if 𝑓(π‘₯, π‘₯, 𝑧) ⊂ 𝐢(𝑔(𝑧)) for each π‘₯ ∈ 𝐸 and 𝑧 ∈
𝑄(π‘₯), then
𝑓 (V, π‘₯, 𝑧) ⊂ΜΈ − int 𝐢 (𝑔 (𝑧)) ,
∀π‘₯ ∈ 𝐸;
∀π‘₯ ∈ 𝐸,
(54)
βŸ¨π‘™ (𝑧) , π‘₯⟩ = ⋃ {βŸ¨π‘, π‘₯⟩ : 𝑝 ∈ 𝑙 (𝑧)} .
(55)
where
The VVIP was investigated in [18, 19]. If 𝑙 is single valued, then
the VVIP becomes a vector variational inequality problem,
which was discussed in [20–22]. Note that 𝑋 is indeed locally
convex.
Assume that 𝐸 is compact and convex and βŸ¨π‘™(π‘₯), π‘₯⟩ is a
singleton for each π‘₯ ∈ 𝐸.
Abstract and Applied Analysis
7
Acknowledgment
Let
𝑋 = 𝑍,
The work was supported by both the Doctoral Fund of
innovation of Beijing University of Technology (2012) and the
11th graduate students Technology Fund of Beijing University
of Technology (No. ykj-2012-8236).
𝐸 = 𝐹,
𝐢 (π‘₯) = 𝐷,
∀π‘₯ ∈ 𝐸,
𝑔 : 𝐸 󳨀→ π‘Œ be any continuous mapping,
𝑓 (π‘₯, 𝑧) = βŸ¨π‘™ (𝑧) , π‘₯⟩,
𝑃 (π‘₯) = 𝐸,
(56)
∀ (π‘₯, 𝑧) ∈ 𝐸 × πΈ,
∀π‘₯ ∈ 𝐸,
𝑄 (π‘₯) = {π‘₯} ,
∀π‘₯ ∈ 𝐸
in Theorem 17. Then the GVQEP1 reduces to the VVIP. All
the assumptions of Theorem 17 are verified as follows.
(a) and (b) Since 𝐷 is a constant cone, π‘₯ 󳨃→ int 𝐷 has a
continuous selection and both 𝐢 and π‘Š are usc.
(c) Obviously, 𝑓 is strict, 𝑔 is continuous, 𝑃 is strict
and continuous, and 𝑄 is strict and usc. In addition, 𝑓 is
continuous with compact values.
In fact, for each fixed (π‘₯0 , 𝑧0 ) ∈ 𝐸 × πΈ, letting π‘š(𝑝) =
βŸ¨π‘, π‘₯0 ⟩, we see that π‘š is continuous with compact values.
So 𝑓(π‘₯0 , 𝑧0 ) = π‘š(𝑙(𝑧0 )) is compact by the compactness of
𝑙(𝑧0 ) and Lemma 3 (2). Moreover, take 𝐿(𝑝, π‘₯, 𝑧) = βŸ¨π‘, π‘₯⟩
and 𝐻(π‘₯, 𝑧) = 𝑙(𝑧). Obviously, 𝐻 is continuous with compact
values. 𝐿 is continuous in view of the bilinearity of itself and
[11, Theorem 2.17]. Thus
𝑓 (π‘₯, 𝑧) =
⋃ 𝐿 (𝑝, π‘₯, 𝑧)
(57)
𝑝∈𝐻(π‘₯,𝑧)
is continuous by [13, Theorem 1].
(d) For each 𝑧 ∈ 𝐸, π‘₯ 󳨃→ 𝑓(π‘₯, 𝑧) is generalized 𝐷quasiconvex by its linearity.
(e) The assertion that 𝑓 has compact values was verified
in (c). Clearly, for each π‘₯ ∈ 𝐸, both 𝑃(π‘₯) and 𝑄(π‘₯) are closed
and convex.
Thus there exist (π‘₯, 𝑧) ∈ 𝐸 × πΈ, π‘₯ ∈ 𝑃(π‘₯) = 𝐸, and 𝑧 ∈
𝑄(π‘₯) = {π‘₯} such that
𝑓 (π‘₯, π‘₯) − 𝑓 (π‘₯, π‘₯) ⊂ΜΈ − int 𝐷,
∀π‘₯ ∈ 𝐸,
that is, βŸ¨π‘™ (π‘₯) , π‘₯⟩ − βŸ¨π‘™ (π‘₯) , π‘₯⟩ ⊂ΜΈ − int 𝐷,
∀π‘₯ ∈ 𝐸.
(58)
Since βŸ¨π‘™(π‘₯), π‘₯⟩ is a singleton for each π‘₯ ∈ 𝐸,
βŸ¨π‘™ (π‘₯) , π‘₯ − π‘₯⟩ ⊂ΜΈ − int 𝐷,
∀π‘₯ ∈ 𝐸,
(59)
which implies that π‘₯ is a solution of VVIP.
Incidentally, the set-valued mapping 𝑙 satisfying the
conditions above exists. For instance, let 𝑋 = R2 , π‘Œ = R,
and
𝐸 = {π‘₯ = (π‘₯1 , π‘₯2 ) : π‘₯12 + π‘₯22 ≤ 1} ⊂ 𝑋.
(60)
Define 𝑙 by
𝑙 (π‘₯) = {(πœ†π‘₯2 , −πœ†π‘₯1 ) : |πœ†| ≤ 1} ,
∀π‘₯ = (π‘₯1 , π‘₯2 ) ∈ 𝐸. (61)
Clearly, 𝑙 is strict and continuous with compact values.
Moreover, βŸ¨π‘™(π‘₯), π‘₯⟩ = {0} for each π‘₯ ∈ 𝐸.
References
[1] X. H. Gong, “Scalarization and optimality conditions for vector
equilibrium problems,” Nonlinear Analysis: Theory, Methods &
Applications A, vol. 73, no. 11, pp. 3598–3612, 2010.
[2] G. Y. Chen, X. Q. Yang, and H. Yu, “A nonlinear scalarization
function and generalized quasi-vector equilibrium problems,”
Journal of Global Optimization, vol. 32, no. 4, pp. 451–466, 2005.
[3] Q. H. Ansari and F. Flores-Bazán, “Recession methods for generalized vector equilibrium problems,” Journal of Mathematical
Analysis and Applications, vol. 321, no. 1, pp. 132–146, 2006.
[4] S. J. Li and P. Zhao, “A method of duality for a mixed vector
equilibrium problem,” Optimization Letters, vol. 4, no. 1, pp. 85–
96, 2010.
[5] G. Y. Chen, “Existence of solutions for a vector variational
inequality: an extension of the Hartmann-Stampacchia theorem,” Journal of Optimization Theory and Applications, vol. 74,
no. 3, pp. 445–456, 1992.
[6] C. Gerth and P. Weidner, “Nonconvex separation theorems
and some applications in vector optimization,” Journal of
Optimization Theory and Applications, vol. 67, no. 2, pp. 297–
320, 1990.
[7] G. Y. Chen and X. Q. Yang, “Characterizations of variable
domination structures via nonlinear scalarization,” Journal of
Optimization Theory and Applications, vol. 112, no. 1, pp. 97–110,
2002.
[8] N. J. Huang, J. Li, and J. C. Yao, “Gap functions and existence of
solutions for a system of vector equilibrium problems,” Journal
of Optimization Theory and Applications, vol. 133, no. 2, pp. 201–
212, 2007.
[9] J. Li and N. J. Huang, “An extension of gap functions for a
system of vector equilibrium problems with applications to
optimization problems,” Journal of Global Optimization, vol. 39,
no. 2, pp. 247–260, 2007.
[10] D. N. Qu and X. P. Ding, “Some quasi-equilibrium problems
for multimaps,” Journal of Sichuan Normal University. Natural
Science Edition, vol. 30, no. 2, pp. 173–177, 2007.
[11] D. T. Luc, Theory of Vector Optimization, vol. 319 of Lecture
Notes in Economics and Mathematical Systems, Springer, Berlin,
Germany, 1989.
[12] W. Rudin, Functional Analysis, International Series in Pure and
Applied Mathematics, McGraw-Hill, New York, NY, USA, 2nd
edition, 1991.
[13] J. P. Aubin and I. Ekeland, Applied Nonlinear Analysis, John
Wiley & Sons, New York, NY, USA, 1984.
[14] L. J. Lin and Z. T. Yu, “On some equilibrium problems for multimaps,” Journal of Computational and Applied Mathematics,
vol. 129, no. 1-2, pp. 171–183, 2001.
[15] J. P. Aubin and A. Cellina, Differential Inclusion, Springer,
Berlin, Germany, 1994.
[16] P. H. Sach and L. A. Tuan, “New scalarizing approach to the
stability analysis in parametric generalized Ky Fan inequality
problems,” Journal of Optimization Theory and Applications,
2012.
8
[17] F. E. Browder, “The fixed point theory of multi-valued mappings
in topological vector spaces,” Mathematische Annalen, vol. 177,
pp. 283–301, 1968.
[18] G. M. Lee, D. S. Kim, B. S. Lee, and S. J. Cho, “Generalized
vector variational inequality and fuzzy extension,” Applied
Mathematics Letters, vol. 6, no. 6, pp. 47–51, 1993.
[19] S. Park, B. S. Lee, and G. M. Lee, “A general vector-valued
variational inequality and its fuzzy extension,” International
Journal of Mathematics and Mathematical Sciences, vol. 21, no.
4, pp. 637–642, 1998.
[20] F. Giannessi, “Theorems of alternative, quadratic programs and
complementarity problems,” in Variational Inequality Complementary Problems, R. W. Cottle, F. Giannessi, and J. L. Lions,
Eds., pp. 151–186, John Wiley & Sons, New York, NY, USA, 1980.
[21] X. Q. Yang, “Vector variational inequality and its duality,”
Nonlinear Analysis: Theory, Methods & Applications A, vol. 21,
no. 11, pp. 869–877, 1993.
[22] X. Q. Yang and C. J. Goh, “On vector variational inequalities:
application to vector equilibria,” Journal of Optimization Theory
and Applications, vol. 95, no. 2, pp. 431–443, 1997.
Abstract and Applied Analysis
Advances in
Operations Research
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Advances in
Decision Sciences
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Mathematical Problems
in Engineering
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Journal of
Algebra
Hindawi Publishing Corporation
http://www.hindawi.com
Probability and Statistics
Volume 2014
The Scientific
World Journal
Hindawi Publishing Corporation
http://www.hindawi.com
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
International Journal of
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Volume 2014
Submit your manuscripts at
http://www.hindawi.com
International Journal of
Advances in
Combinatorics
Hindawi Publishing Corporation
http://www.hindawi.com
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Journal of
Complex Analysis
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
International
Journal of
Mathematics and
Mathematical
Sciences
Journal of
Hindawi Publishing Corporation
http://www.hindawi.com
Stochastic Analysis
Abstract and
Applied Analysis
Hindawi Publishing Corporation
http://www.hindawi.com
Hindawi Publishing Corporation
http://www.hindawi.com
International Journal of
Mathematics
Volume 2014
Volume 2014
Discrete Dynamics in
Nature and Society
Volume 2014
Volume 2014
Journal of
Journal of
Discrete Mathematics
Journal of
Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com
Applied Mathematics
Journal of
Function Spaces
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Optimization
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Download