Acta Mathematicae Applicatae Sinica, English Series
Vol. 30, No. 4 (2014) 845–858
DOI: 10.1007/s10255-014-0426-4
http://www.ApplMath.com.cn & www.SpringerLink.com
Acta Mathemacae Applicatae Sinica,
English Series
© The Editorial Office of AMAS &
Springer-Verlag Berlin Heidelberg 2014
Painlevé-Kuratowski Convergences of the Solution Sets
for Perturbed Vector Equilibrium Problems without
Monotonicity
Zai-yun PENG1 , Xin-min YANG2
1 College
of Science, Chongqing JiaoTong University, Chongqing 400074, China
(E-mail: pengzaiyun@126.com)
2 Department
of Mathematics, Chongqing Normal University, Chongqing 400047, China
(E-mail: xmyang@cqnu.edu.cn)
Abstract
In this paper, we obtain some stability results for perturbed vector equilibrium problems. Under
new assumptions, which are weaker than the assumption of C-strict monotonicity, we provide sufficient conditions
for the Painlevé-Kuratowski Convergence of the weak efficient solution sets and efficient solution sets for the
perturbed vector equilibrium problems with a sequence of mappings converging in real linear metric spaces.
These results extend and improve some known results in the literature.
Keywords
stability; Painlevé-Kuratowski convergence; efficient solution; perturbed vector equilibrium prob-
lem; scalarization
2000 MR Subject Classification
1
90C33; 49K40; 34D10
Introduction
It is well known that the vector equilibrium problem (VEP) is a very general mathematical
model, which embraces the formats of several disciplines, as those for equilibrium problems of
mathematical physics, Game Theory, vector variational inequality problem, (vector) complementarity problem and (vector) saddle point problem, and so on (see [12,13]). In the literature,
existence results for various types of vector equilibrium problems have been investigated intensively (see [9,12,13,14] and references therein).
The stability analysis of the solution set maps for parametric (VEP) is of considerable interest. Some results on the semicontinuity of the solution set maps for the parametric (VEP) (or
parametric variational inequalities) with the parameter perturbed in the space of parameters
are now available in the literature. Cheng and Zhu[7] discussed the upper semicontinuity and
the lower semicontinuity of the solution mapping for a class of parametric variational inequalities in finite-dimensional spaces. Gong and Yao[16] first obtained the lower semicontinuity of the
efficient solutions of parametric generalized systems by virtue of a density result and scalarization technique. By using the ideas of Cheng and Zhu[7] , Gong[17] discussed the continuity of the
solution set mapping for a class of parametric weak vector equilibrium problems in topological
vector spaces. Huang et al.[19] discussed the parametric implicit vector equilibrium problems
and established the lower semicontinuity of solution mapping using the local existence results of
the solutions. Anh and Khanh[3] investigated the semicontinuity of solution mapping of paraManuscript received July 15, 2013. Revised October 9, 2013.
Supported by the National Natural Science Foundation of China (No.11301571.11271389.11271391), the Natural
Science Foundation Project of ChongQing (No.CSTC, 2012jjA00016.2011BA0030) and the Education Committee
Research Foundation of ChongQing (KJ130428).
Z.Y. PENG, X.M. YANG
846
metric vector quasi-equilibrium problems by virtue of the closedness or openness assumptions
for some certain sets. Chen and Li[6] discussed the continuity of various efficient solution sets
for a parametric generalized system and improved the results of [16,17]. Recently, Peng and
Yang[27] obtained sufficient conditions for the lower semicontinuity of the solution mappings for
two classes of parametric weak generalized equilibrium problems (they called weak generalized
Ky Fan Inequalities) in Hausdorff topological vector spaces under weak C-strict monotonicity,
which improved the results of [6,16].
As for the stability result investigated on the convergence of the sequence of mappings, there
are some results for the vector optimization and vector variational inequality with a sequence
of sets converging in the sense of Painlevé-Kuratowski (e.g., [10, 20-22]). Huang[20] discussed
the convergence of the approximate efficient sets to the efficient sets of vector-valued and setvalued optimization problems in the sense of Painlevé-Kuratowski and Mosco. Lucchetti and
Miglierina[22] investigated the Painlevé-Kuratowski set convergence of the solution sets of the
perturbed problems both in the given space and its image space for a convex vector optimization
problem. But, to the best of our knowledge, there are few stability results available for the
perturbed vector equilibrium problem with the convergence of a sequence of mappings. It
appears that two relevant paper are [9] and [11]. In [9], where Durea considered the vector
equilibrium problems with the perturbations of the multifunction and obtained the PainlevéKuratowski upper convergence of the solution sets. In [11], under the C-strict monotonicity,
Fang et al. obtained the Painlevé-Kuratowski Convergence of the efficient solution sets, the
weak efficient solution sets and various proper efficient solution sets for a perturbed generalized
system with a sequence of mappings converging in locally convex Hausdorff topological vector
spaces.
Since the perturbed vector equilibrium problem with a sequence of mappings converging
is different from the parametric vector equilibrium problem with the parameter perturbed in
a space of parameters, it is important to study the Painlevé-Kuratowski Convergence of the
sequence of the solution sets. We also observe that the Painlevé-Kuratowski Convergence of
the solution sets to perturbed vector equilibrium problem has been discussed under the restrictive assumption of C-strict monotonicity, which implies that the f -solution set of the perturbed vector equilibrium problem is a singleton for a linear continuous functional f (see recent
literature[11] ). However, it is well known that the f -solution set of the perturbed vector equilibrium problem should be general, but not a singleton. So, in this paper, we aim at studying the
Painlevé-Kuratowski Convergence of the solution sets to perturbed vector equilibrium problem
when the f -solution set is a general set by removing the assumption of C-strict monotonicity.
The obtained results extend and improve the recent ones in the literature[11] .
The rest of the paper is organized as follows. In Section 2, we introduce the problems (VEP)
and (VEP)n , recall some definitions and important properties. In Section 3 and Section 4, we
discuss the Painlevé-Kuratowski convergences of the weak efficient solution sets and the efficient
solution sets, respectively, and provide some examples to illustrate that our main results extend
the corresponding ones in [11,17,27].
2
Preliminaries
Throughout this paper, unless otherwise specified, d(·, ·) denote the metric in any metric space.
Let B(0, δ) denote the closed ball with radius δ ≥ 0 and center 0 in any metric linear spaces.
Let X and Y be two real linear metric spaces. Let Y ∗ be the topological dual space of Y , and
C be a closed convex pointed cone in Y with nonempty topological interior intC.
Let
C ∗ := f ∈ Y ∗ : f (y) ≥ 0, ∀ y ∈ C
Convergences of the Solution Sets for Perturbed Vector Equilibrium Problems without Monotonicity
847
be the dual cone of C. Denote the quasi-interior of C ∗ by C , i.e,
C := {f ∈ Y ∗ : f (y) > 0, ∀y ∈ C \ {0} .
It is easy to see that C = ∅ if and only if C has a base.
Let A be a nonempty subset of X and F : A × A → Y be a bifunction. We consider the
following vector equilibrium problem
(VEP) Find x ∈ A such that F (x, y) ∈ −K,
∀ y ∈ A,
where K ∪ {0} is a convex cone in Y .
For a sequence of bifunctions Fn : A × A → Y, n = 1, 2, · · · , we define a sequence of vector
equilibrium problems
(VEP)n
Find xn ∈ A such that Fn (xn , y) ∈ −K,
∀ y ∈ A,
where K ∪ {0} is a convex cone in Y .
Special Case.
Let X = E and Y = E ∗ , where E is a reflexive Banach space with dual E ∗ . Let An be a
single-valued operators from E to E ∗ . Let Fn (x, y) = An (x), y − x , where ·, · denotes the
inner product. Then, (VEP)n reduces to (VI)n considered in [21].
For each f ∈ C ∗ \ {0}, let Vf and Vfn denote the set of f -efficient solutions to the (VEP)
and (VEP)n , i.e.,
Vfn = x ∈ A : f (Fn (xn , y)) ≥ 0, ∀ y ∈ A .
Vf = x ∈ A : f (F (x, y)) ≥ 0, ∀ y ∈ A ,
Throughout this paper, for all f ∈ C ∗ \ {0}, we always assume Vf = ∅ and Vfn = ∅.
In this paper, under new assumptions, we discuss the Painlevé-Kuratowski convergence of
the weak efficient solution sets and the efficient solution sets of (VEP)n .
In the following, we recall some concepts of the convergence of set sequence and mapping
sequence which will be used in the sequel.
Definition 2.1[9,10] . Let X be a normed space. A sequence of sets {An ⊂ X : n ∈ N } is said
P.K.
to converge in the sense of Painlevé-Kuratowski (P.K.) to A(denoted as An −→ A) if
lim sup An ⊂ A ⊂ lim inf An
n→∞
n→∞
with
lim inf An := x ∈ X | ∃ (xn ), xn ∈ An , ∀ n ∈ N, xn → x ,
n→∞
lim sup An := x ∈ X | ∃ (nk ), ∃ (xnk ), xnk ∈ Ank , ∀ k ∈ N, xnk → x .
n→∞
Definition 2.2[26] .
Let Fn , F : X → Y (n ∈ N ) be a vector-valued mapping and let U (x) be
Γ
C
F)
the family of neighborhoods of x. We say that (Fn )n∈N ΓC -converges to F (denoted as Fn −→
if for every x ∈ X :
(i) ∀ U ∈ U (x), ∀ ε ∈ int C, ∃nε,U ∈ N such that ∀ n ≥ nε,U , ∃ xn ∈ U such that
Fn (xn ) ∈ F (x) + ε − C;
(ii) ∀ ε ∈ int C, ∃ Uε ∈ U (x), kε ∈ N such that ∀ x ∈ Uε , ∀ n ≥ kε ,
Fn (x ) ∈ F (x) − ε + C.
Z.Y. PENG, X.M. YANG
848
Definition 2.3[28] . Let Fn , F : X → Y (n ∈ N ) be vector-valued mapping. We say that Fn
continuously converges to F if the fact that xn → x implies that Fn (xn ) → F (x).
Let T : Λ → 2X be a set-valued mapping, and given λ ∈ Λ.
Definition 2.4[1,2].
(i) T is called lower semicontinuous (l.s.c, in short) at λ iff for any open set V satisfying
V T (λ) = ∅, there exists δ > 0, such that for every λ ∈ B(λ, δ), V ∩ T (λ) = ∅.
(ii) We say T is l.s.c on Λ iff it is l.s.c at each λ ∈ Λ.
Definition 2.5. Let F : X × X → Y be a vector-valued mapping.
(i) F (x, ·) is called C-convex if, for each x1 , x2 ∈ A and t ∈ [0, 1], tF (x, x1 ) + (1 −
t)F (x, x2 ) ∈ F (x, tx1 + (1 − t)x2 ) + C.
(ii) F (x, ·) is called C-convexlike on A, iff for any x1 , x2 ∈ A and any t ∈ [0, 1], there exists
x3 ∈ A such that tF (x, x1 ) + (1 − t)F (x, x2 ) ∈ F (x, x3 ) + C.
(iii) A set D ⊂ Y is called a C-convex set, iff D + C is a convex set in Y.
Remark 2.1. (i) F is C-convexlike on X if and only if F (X) + C is convex.
(ii) From the definitions, we obtain immediately the following implications for the map F :
C-convexity ⇒ C-convexlikeness.
However, one simple example in [25] (Example 3.2) shows that the converse implication is
generally not valid. Hence, the class of C-convexlike maps is larger than the class of C-convex
maps.
Proposition 2.1[28] . For any sequence of sets C ν , Dν ⊂ Rn , both the inner limit set lim inf C ν
ν
and the outer limit set lim sup C ν are closed. Furthermore, they depend only on the closures cl
C ν in the sense that
ν
cl C ν = clDν =⇒
⎧
⎨ lim inf C ν = lim inf Dν ,
ν
ν
⎩ lim sup C ν = lim sup Dν .
ν
ν
Thus, whenever lim C ν exists, it is closed. (If C ν ≡ C, then lim C ν = cl C).
ν
ν
[2,4]
Proposition 2.2
. Let X and Y be topological spaces, T : X → 2Y be a set-valued mapping.
T is l.s.c. at x0 ∈ X if and only if for any net {xα } ⊂ X with xα → x0 and any y0 ∈ T (x0 ),
there exists yα ∈ T (xα ) such that yα → y0 .
3
Painlevé-Kuratowski Convergence of the Weak Efficient
Solution Sets
Denote by I W and InW the sets of the weak efficient sets to (VEP) and (VEP)n , respectively,
i.e.,
I W = x ∈ A | F (x, y) ∈ Y \ −int C, ∀ y ∈ A ,
InW = xn ∈ A | Fn (xn , y) ∈ Y \ −int C, ∀ y ∈ A .
In this section, without using the assumption of C-strict monotonicity, we mainly discuss
the Painlevé-Kuratowski Convergence of InW .
Lemma 3.1. Let A be a nonempty compact set. For n = 0, 1, 2 · · · , assume that the following
conditions are satisfied:
Convergences of the Solution Sets for Perturbed Vector Equilibrium Problems without Monotonicity
849
Γ
C
(i) For each y ∈ A, Fn (·, y) −→
F (·, y);
Γ
C
−F (·, y).
(ii) For each y ∈ A, −Fn (·, y) −→
Then, for each f ∈ C ∗ \ {0}, lim Vfn = Vf .
n→∞
Proof. 1) “ lim Vfn
n→∞
any xn ∈ Vfn with xn
⊃ Vf ”. On the contrary, assume that there exists x0 ∈ Vf , such that for
x0 .
Since x0 ∈ A and A is compact, there exists a net x̂n ∈ A such that x̂n → x0 . Obviously,
x̂n ∈
/ Vfn . Then there exists yβ ∈ A such that f (Fn (x̂n , yβ )) < 0. Take an arbitrary subnet
{x̂nk } ⊆ {x̂n } ⊆ A (taking a subnet of {x̂nk } if necessary), then x̂nk → x0 . Also, we have
f (Fnk (x̂nk , yβ )) < 0.
(3.1)
It follows from x0 ∈ Vf and yβ ∈ A that
f (F (x0 , yβ )) ≥ 0.
(3.2)
Γ
C
F (·, yβ ), it follows from Definition 2.2 that for any ε ∈ int C, ∃ nε ∈ N such
Since Fn (·, yβ ) −→
that
Fnk (xnk , yβ )) ∈ F (x0 , yβ )) − ε + C,
∀ nk ≥ n ε .
Furthermore, we have f (Fnk (xnk , yβ ))) ≥ f (F (x0 , yβ )) − f (ε).
From the arbitrariness of ε, we get
f (Fnk (xnk , yβ ))) ≥ f (F (x0 , yβ )).
(3.3)
Combining with (3.2) and (3.3), we obtain f (Fnk (xnk , yβ ))) ≥ 0, which contradicts (3.1). Then
for each x ∈ Vf , there exists xn ∈ Vfn such that xn → x (n → ∞), i.e., x ∈ lim Vfn . So we
n→∞
conclude lim Vfn ⊃ Vf .
n→∞
nq
2) “ lim Vfn ⊂ Vf ”. Take any x ∈ lim Vfn . Then there exists xnq ∈ Vf
n→∞
n→∞
such that
lim xnq = x, where {nq } is a subnet of N . Then, we have that for each y ∈ A
n→∞
f (Fnq (xnq , y)) ≥ 0.
(3.4)
Γ
C
−F (·, y), we get that for any ε ∈ int C, ∃ lε ∈ N such that
Because −Fn (·, y) −→
−Fnq (xnq , y)) ∈ −F (x, y) − ε + C,
∀ nq ≥ l ε .
So we conclude f (Fnq (xnq , y)) − f (F (x, y)) − f (ε) ≤ 0.
From the arbitrariness of ε, we get
f (F (x, y)) ≥ f (Fnq (xnq , y)),
∀ y ∈ A.
(3.5)
By (3.4) and (3.5), we can obtain f (F (x, y)) ≥ 0 for each y ∈ A, i.e. x ∈ Vf . This completes
the proof.
2
Remark 3.1. In [11], under the condition of C-strict monotonicity, the relation of Vfn and
Vf to (VEP)n and (VEP) are obtained (Lemma 2.2 of [11]). However, this condition is so
strict that the f -effcient solution set to (VEP)n and (VEP) is confined to be a singleton. In
our paper, we discuss the relationships of Vfn and Vf to (VEP)n and (VEP) without using
C-strict monotonicity. Furthermore, the f -efficient solution set may be a general set, but not
a singleton. The following example is given to illustrate the case.
Z.Y. PENG, X.M. YANG
850
2
Example 3.1. Let X = R, Y = R2 , C = R+
= x = (x1 , x2 ) : x1 ≥ 0, x2 ≥ 0 . Let
A = [0, 1]. It is clear that A is a nonempty, compact set.
Define the mapping Fn : A × A → R2 by
3
1
3
x−
, n = 0,
for all x, y ∈ A,
F (x, y) = − (x2 + y 2 ),
2
4
8
3
1 2
1
1
1 2 3
Fn (x, y) = −
x−
x−
−
,
,
+ y−
2
n
n
4
n
8
n = 1, 2 · · · ,
for all x, y ∈ A.
It is clear that A is compact. Let f = (0, 1) ∈ C ∗ \ {0}. Then, it is easy to get that
= 18 + n1 , n1 + 1 and Vf = 18 , 1 . Obviously, they are general set, but not a singleton.
Because the verification method for Condition (i) is similar to (ii), now we check Condition
(ii) of Lemma 3.1 as follows:
(a) ∀ ε ∈ int C, ∃ Uε = x− n1 , x+ n1 ∈ U (x), ∃N such that ∀ x1 ∈ Uε = x− n1 , x+ n1 , ∀ n ≥
N, we have
3
1
3
1 2
1 2
1
x1 −
x1 −
−
,−
−Fn (x1 , y) =
+ y−
2
n
n
4
n
8
3
2y
2
1
3
2x1
1
x21 + y 2 −
−
+ 2 , − x1 − −
,
=
2
n
n
n
4
n 8
3
3
2
2
3
−Fn (x1 , y) + F (x, y) =
(x21 − x2 ) − (x1 + y) + 2 , − (x1 − x) +
.
2
n
n
4
4n
Vfn
Then we obtain
−Fn (x1 , y) ∈ −F (x, y) − ε + C.
(b) ∀ U ∈ U (x), ∀ ε ∈ int C, ∃ nε,U ∈ N such that ∀ n ≥ nε,U , ∃ xn = x +
−Fn (xn , y) + F (x, y) =
3y 1
2n n
− 2y , −
1
n
∈ U such that
3
.
32
Thus, we also get
−Fn (xn , y) ∈ −F (x, y) + ε − C.
So, the conditions are fulfilled. It follows from Lemma 3.1 that, for each f ∈ C ∗ \{0}, lim Vfn =
n→∞
Vf = 18 , 1 .
However, the condition that Fn is a C-strictly monotone mapping is violated. Indeed,
x = 12 , y = 1 ∈ A and x = y, we find that
F0 (x, y) + F0 (y, x) =
1
3
(x + y) −
4
4
∈ −intC
2 1
3
(x+y)− −
4
n 4
∈ −intC,
− 3(x2 + y 2 ),
and
1
Fn (x, y)+Fn (y, x) = −3 x−
n
2
1
+ y−
n
2
,
for n ≥ 2.
Hence, the Lemma 2.2 of [11] is not applicable.
In a similar way of Lemma 2.1 in [14], we obtain the following Lemma.
Lemma 3.2.
on A. Then,
Suppose that int C = ∅ and for n = 0, 1, 2, · · · , ∀ x ∈ A, Fn (x, ·) is C-convexlike
InW =
f ∈C ∗ \{0}
Vfn .
Convergences of the Solution Sets for Perturbed Vector Equilibrium Problems without Monotonicity
851
Theorem 3.3. Let A be a nonempty compact set. For n = 0, 1, 2 · · · , assume that the following conditions are satisfied:
ΓC
F (·, y);
(i) For each y ∈ A, Fn (·, y) −→
Γ
C
−F (·, y).
(ii) For each y ∈ A, −Fn (·, y) −→
(iii) For each x ∈ A, Fn (x, ·) is C-convexlike on A.
Then, we have InW → I W in the sense of Painlevé-Kuratowski.
Proof.
i) First, we prove lim sup InW ⊆ I W . On the contrary, assume that there exists x ∈
n→∞
lim sup InW with x ∈ I W .
n→∞
From x ∈ lim sup InW , we have x = lim xnk , where xnk ∈ InWk and {nk } is a subnet of N.
Thus, we have
nk →∞
n→∞
Fnk (xnk , y) ∈ Y \ −int C,
∀ y ∈ A.
(3.6)
W
Noting that x ∈ I , there exists y0 ∈ A such that
F (x, y0 ) ∈ −int C.
(3.7)
In particular, for (3.6), we have
Fnk (xnk , y0 ) ∈ Y \ −int C.
Γ
C
Since −Fn (·, y0 ) −→
−F (·, y0 ), we have ∀ ε ∈ int C, ∃ Uε ∈ U (x), kε ∈ N such that for
xnk ∈ Uε , ∀ n ≥ kε ,
−Fn (xnk , y0 ) ∈ −F (x, y0 ) − ε + C.
Then,
F (x, y0 ) ∈ Y \ −int C,
which contradicts (3.7). Hence, lim sup InW ⊆ I W .
n→∞
ii) We prove I W ⊆ lim inf InW . By virtue of Lemma3.2, we need to prove that
n→∞
f ∈C ∗ \{0}
Vf ⊂ lim inf
n→∞
Vfn .
f ∈C ∗ \{0}
Vf , from Lemma 3.1, there exists f ∈ C ∗ \ {0} such that x ∈
Vfn , such that xn → x. This implies
Vf = lim Vfn . Then, there exists xn ∈ Vfn ⊂
n→∞
∗
f ∈C \{0}
Vfn . Therefore, we have
Vf ⊂ lim inf
Vfn . The proof is
that x ∈ lim
In fact, for any x ∈
f ∈C ∗ \{0}
n→∞ f ∈C ∗ \{0}
f ∈C ∗ \{0}
complete.
2
Example 3.2. Let X = R, Y = R2 , C = R+
A = [0, 1]. It is clear that A is a compact set.
Define the mapping Fn : A × A → R2 by
n→∞ f ∈C ∗ \{0}
2
= x = (x1 , x2 ) : x1 ≥ 0, x2 ≥ 0 . Let
1 2
,
n = 0, for all x, y ∈ A,
F (x, y) = ey (siny − x2 + 1), sin(y − 1) + x −
2
1 2
1
Fn (xn , y) = ey (siny − x2n + 1), sin(y − 1) + xn −
, n = 1, 2 · · · , where xn = x − .
2
n
Using the same method as that in Example 3.1, it is easy to check that Conditions (i)–(ii)
of Theorem 3.3 are satisfied.
Z.Y. PENG, X.M. YANG
852
The condition (iii) of Theorem 3.3 can be checked as follows: Let g(y) = ey (siny − x2n +
1), h(y) = sin(y − 1) + (xn − 12 )2 , we have g (y) = ey (siny − x2n + 1) + ey cos, g (y) =
ey (2cosy − x2n + 1) > 0, h (y) = cos(y − 1), h (y) = −sin(y − 1) ≥ 0. Obviously, for each
x ∈ A, F (x, ·) is C-convexlike on A. So, by virtu of Theorem 3.3, InW −→ I W in the sense of
Painlevé-Kuratowski.
4
Painlevé-Kuratowski Convergence of the Efficient Solution Sets
Denote by I and In the sets of the efficient sets to (VEP) and (VEP)n , respectively, i.e.,
I = x ∈ A | F (x, y) ∈
/ −C \ {0}, ∀ y ∈ A ,
In = xn ∈ A | Fn (xn , y) ∈
/ −C \ {0}, ∀ y ∈ A .
In this section, without using C-strict monotonicity, we will mainly discuss the PainlevéKuratowski Convergence of the efficient solution sets of (VEP)n .
Definition 4.1[14] . A vector x ∈ A is called a positive proper efficient solution to the (VEP)
if there exists f ∈ C such that
f (F (x, y)) ≥ 0,
∀ y ∈ A.
First, we give a result of lower semicontinuous to Vfn for (VEP)n .
Lemma 4.1. Let f ∈ C ∗ \ {0}. Suppose the following conditions are satisfied:
(i) A is a nonempty, compact set;
(ii) Fn (·, ·) continuously converges to F (·, ·) on X × X;
(iii) For n = 0, 1, 2, · · · , each x ∈ A \ Vfn , there exists y ∈ Vfn , such that
Fn (x, y) + Fn (y, x) + B(0, dr (x, y)) ⊂ −C,
where γ > 0 is a positive constant.
Let us define the set-valued mapping Hn : C ∗ \ {0} → 2A by
Hn (f ) = Vfn ,
∀ f ∈ C ∗ \ {0}.
then we have Hn (·) is l.s.c on C ∗ \ {0}.
Proof. We just give the proof of the case when n = 0, that is, H(·) is l.s.c on C ∗ \ {0} (set
H0 (·) = H(·)). Suppose to the contrary that there exists f0 ∈ C ∗ \ {0}, such that H(·) be
not l.s.c. at f0 . Then, there exist a sequence {fm } with fm → f0 with respect to the topology
β(Y ∗ , Y ) and x0 ∈ H(f0 ) = Vf0 such that for any xm ∈ H(fm ) = Vfm , xm x0 .
Since x0 ∈ A and A is nonempty compact, then there exists xm ∈ A, such that xm → x0 .
Obviously, xm ∈ A \ H(fm ). By (iii), there exists ym ∈ H(fm ) such that
F (xm , ym ) + F (ym , xm ) + B(0, dr (xm , ym )) ⊂ −C,
(4.1)
where γ > 0 is a positive constant. For ym ∈ H(fm ) implies ym ∈ A, because A is nonempty
compact, there exist y0 ∈ A and a subsequence {ymk } of {ym }, such that ymk → y0 . In
particular, for (4.1), we have
F (xmk , ymk ) + F (ymk , xmk ) + B(0, dr (xmk , ymk )) ⊂ −C.
(4.2)
Taking the limit as mk → +∞, it follows from the assumption (ii) that we have
F (x0 , y0 ) + F (y0 , x0 ) + B(0, dr (x0 , y0 )) ⊂ −C.
(4.3)
Convergences of the Solution Sets for Perturbed Vector Equilibrium Problems without Monotonicity
853
Assume that x0 = y0 , by (4.3), we obtain F (x0 , y0 ) + F (y0 , x0 ) ∈ −int C. Thus, it follows
from f0 ∈ C ∗ \ {0} that we have
f0 (F (x0 , y0 ) + F (y0 , x0 )) < 0.
(4.4)
Noting that x0 ∈ H(f0 ) and y0 ∈ A, we have
f0 (F (x0 , y0 )) ≥ 0.
(4.5)
Moreover, since ymk ∈ H(fmk ) and xmk ∈ A, it follows from the continuity of f0 and assumption
(ii) that
f0 (F (y0 , x0 )) ≥ 0.
(4.6)
By (4.5), (4.6) and the linearity of f0 , we have
f0 (F (x0 , y0 ) + F (y0 , x0 )) ≥ 0.
which contradicts (4.4). Therefore x0 = y0 , which also leads to a contradiction. Indeed, if
x0 = y0 , for ymk ∈ H(fmk ), it has ymk → y0 = x0 , this contradicts that for any xmk ∈ H(fmk ),
xmk does not converge to x0 . Thus, H(·) is l.s.c on C ∗ \ {0}. This completes the proof.
2
Lemma 4.2. Let f ∈ C ∗ \ {0} and C = ∅. Suppose the following conditions are satisfied:
(i) A is a nonempty, compact set;
(ii) Fn (·, ·) continuously converges to F (·, ·) on X × X;
(iii) For n = 0, 1, 2, · · · , each x ∈ A \ Vfn , there exists y ∈ Vfn , such that
Fn (x, y) + Fn (y, x) + B(0, dr (x, y)) ⊂ −C,
where γ > 0 is a positive constant;
(iv) For n = 0, 1, 2, · · · , each x ∈ A, Fn (x, ·) is C-convexlike on A.
Then,
Vfn ⊂ In ⊂ cl
Vfn .
f ∈C
Proof.
f ∈C
We just give the proof of the case when n = 0, that is,
Vf ⊂ I ⊂ cl
Vf .
f ∈C
f ∈C
Since Vf = ∅ for each f ∈ C ∗ \ {0}. Then, by definition, we have
Vf ⊂ I ⊂ I w .
(4.7)
f ∈C
Since for any x ∈ A, F (x, ·) is C-convexlike, then F (x, A) + C is a convex set. From Lemma
3.2, we have
Iw =
Vf .
(4.8)
f ∈C ∗ \{0}
By (4.7) and (4.8), we get
f ∈C
Vf ⊂ I ⊂
f ∈C ∗ \{0}
Vf .
(4.9)
Z.Y. PENG, X.M. YANG
854
We need to show that
Vf ⊂ cl
f ∈C ∗ \{0}
Vf .
f ∈C
Let us define set-valued mapping Hn : C ∗ \ {0} → 2A by
Hn (f ) = Vfn ,
∀ f ∈ C ∗ \ {0}.
∗
From Lemma 4.1,
Hn (f ) is lower semicontinuous on∗ C \ {0}.
Let x0 ∈
Vf . Then, there exists f0 ∈ C \ {0} such that
f ∈C ∗ \{0}
x0 ∈ Vf0 = H( f0 ).
Since C = ∅, let g ∈ C and set
fm = f0 + (1/m)g.
Then, fm ∈ C . We show that {fm } converges to f0 with respect to the weak∗ topology
β(Y ∗ , Y ).
For any neighborhood U of 0 with respect to β(Y ∗ , Y ), there exist bounded subsets Bi ⊂
Y (i = 1, 2, · · · , k) and ε > 0 such that
k
f ∈ Y ∗ : sup |f (y)| < ε ⊂ U.
y∈Bi
i=1
Since Bi is bounded and g ∈ Y ∗ , |g(Bi )| is bounded for i = 1, · · · , k. Thus, there exists K such
that
sup |(1/m)g(y)| < ε,
i = 1, · · · , k, k ≥ K.
y∈Bi
Hence (1/m)g ∈ U, that is, fm − f0 ∈ U. This means that {fm } converges to f0 with respect to
β(Y ∗ , Y ).
Since H(f ) is l.s.c at f0 (set H0 (f ) = H(f )), then
for sequence {fm } ⊂ C ∗ \ {0}, fm → f0
and x0 ∈ H(f0 ), there exists xm ∈ H(fm ) = Vfm ⊂
Vf , such that xm → x0 . This means
that
x0 ∈ cl
By the arbitrariness of x0 ∈
Vf .
f ∈C
f ∈C ∗ \{0}
Vf , we have
f ∈C ∗ \{0}
By (4.9) and (4.10), we obtain that
f ∈C
This completes the proof.
f ∈C
Vf ⊂ cl
Vf .
(4.10)
f ∈C
Vf ⊂ I ⊂ cl
Vf .
f ∈C
2
Remark 4.1. Lemma 4.2 improves and extends Lemma 4.1 of [11] and Lemma 1.2 of [16].
In [11] and [16], the density results have been presented under the condition of C-strict monotonicity (called C-strong monotonicity in [16]), where the f -solution set for (VEP) and (VEP)n
is confined to be a singleton (also can see Lemma 1.1 of [16], Lemma 4.1 of [17]). In our paper,
Convergences of the Solution Sets for Perturbed Vector Equilibrium Problems without Monotonicity
855
we use Assumption (iii) of Lemma 4.2 to weaken this condition. Furthermore, the f -efficient (or
positive proper efficient) solution set may be a general set, but not a singleton. The following
example is given to illustrate the case.
2
Example 4.1. Let X = R, Y = R2 , C = R+
:= [0, +∞) × [0, +∞), A = [−1, 1]. For each
2
x, y ∈ A, define the mapping Fn : A × A −→ R by
1
1
8
2
3
n = 0,
for all x, y ∈ A,
F (x, y) = − x2 + x + y − , 4x 3 y 2 + x2 ,
5
2
3
2
1
8
1 13 2 3
1 2
1
1 2 2
x − 2 + y − ,4 x − 2
x− 2
y +
,
Fn (x, y) = − x − 2 +
n
5
n
2
3
n
2
n
n = 1, 2 · · · ,
for all x, y ∈ A.
For any given μ > 0, let f ((x, y)) = μ1 y. It follows from a direct computation that
= n12 , 1 , Vf (A, F ) = [0, 1]. Obviously, the f -solution set of (GS) and (GS)n is
set-valued, but not a singleton.
It is clear that conditions (i), (ii) and (iv) of Lemma 4.2 are satisfied. The assumption (iii)
can be checked as follows (we just give the verification of the case when n = 0):
For any x ∈ A \ Vf (A, F ) = [−1, 0), there exists y = 0 ∈ Vf (A, F ) and r = 73 > 0 such that
Vfn (A, F )
F (x, y) + F (y, x) + B(0, dr (x, y))
1
1
8
2
3
= − x2 + x + y − , 4x 3 y 2 + x2
5
2
3
2
1
1
8
2
3
2
2
+ − y + y + x − , 4y 3 x + y 2 + B(0, dr (x, 0))
5
2
3
2
9
7
2
r
= − x + x − 8, 6x 3 + B(0, d (x, 0)) ⊂ −C.
10
By Lemma 4.2, we obtain
Vfn ⊂ In ⊂ cl
Vfn .
f ∈C
f ∈C
However, the condition of C-strict monotonicity in [17, 18] (or called C-strong monotonicity)
does not hold. Indeed, for any x ∈ A \ Vf (A, F ) = [−1, 0), there exists y = −x ∈ Vf (A, F ) =
[0, 1], such that
F (x, y) + F (y, x)
1
1
8
2
3
= − x2 + x + y − , 4x 3 y 2 + x2
5
2
3
2
=(−2x2 − 8, 0) ∈ −∂C,
+
1
1
8
2
3
− y 2 + y + x − , 4y 3 x2 + y 2
5
2
3
2
where ∂C is the boundary of C. Obviously, F (x, y) + F (y, x) ∈ −int C, which implies F (·, ·) is
not C-strictly monotone on A × A. Then, the density Lemma 4.1 in [11] and Lemma 1.2 of [16]
are not applicable.
Theorem 4.3. Suppose that all conditions of Lemma 4.2 are satisfied. Then, we have In → I
in the sense of Painlevé-Kuratowski.
Proof. i) “lim sup In ⊂ I”. First, we prove lim sup
Vfn ⊂
Vf . Take any x ∈
n→∞
n→∞ f ∈C
f
∈C
Vfn . Then, there exist xnk ∈
Vfnk such that
lim sup
n→∞ f ∈C
f ∈C
x = lim xnk ,
k→∞
where {nk } is a subset of N . Therefore, there exists f ∈ C such that xnk ∈ Vfn k . Using the
same method of Lemma 3.1, we obtain
x = lim xnk ∈ lim Vfn k = Vf ,
nk →∞
nk →∞
Z.Y. PENG, X.M. YANG
856
then we have x ∈
f ∈L
Vf , i.e.,
lim sup
n→∞
Vfn ⊂
f ∈C
Vf .
By virtue of Lemma 4.2, we have
Vfn ⊂ In ⊂ cl
Vfn ,
f ∈C
and
(4.11)
f ∈C
n = 1, 2, · · ·
(4.12)
f ∈C
Vf ⊂ I ⊂ cl
f ∈C
Vf ,
n = 0.
(4.13)
f ∈C
From (4.12) and (4.13), we have cl (In ) = cl
f ∈C
Vfn ). Then, by Proposition 2.1, we get
lim sup In = lim sup
n→∞
n→∞
(4.14)
f ∈C
It follows from (4.11), (4.13), (4.14) that lim sup In ⊂
n→∞
Vfn .
f ∈C
Vf ⊂ I.
ii) Using the same method of i), with suitable modification, we can get that I ⊂ lim inf In .
P.K.
n→∞
Hence, we obtain that lim sup In ⊂ I ⊂ lim inf In , i.e., In −→ I. This completes the proof.
n→∞
n→∞
2
Remark 4.2. Theorem 4.3 generalizes and improves the corresponding results of [11] (Theorems 4.1–4.2) in the following three aspects:
(i) The condition that A(·) is convex valued is removed;
(ii) The C-strict monotonicity is substituted by Assumption (iii), then the f -efficient solution may be general in our paper;
(iii) The condition that C-convexity of Fn (A, A) is extended to C-convexlikeness;
(iv) The boundness condition for Fn (x, x) ∈ C (∀ x ∈ A) and Fn (A, A) are removed.
We give an example to illustrate that our results extend the corresponding results in the
literature.
2
Example 4.2. Let X = R, Y = R2 , C = R+
= x = (x1 , x2 ) : x1 ≥ 0, x2 ≥ 0 . Let
A = [0, 1]. It is clear that A is a compact set.
Define the mapping Fn : A × A −→ R2 by
9
3
F (x, y) = ey (sin(x − 1) − 1) + (x + y) − ,
2
2
11 2 1
x y + x+3 ,
n = 0,
for all x, y ∈ A,
3
2
1
1
3
9
Fn (x, y) = ey sin x − 3 − 1 − 1 +
x− 3 +y − ,
n
2
n
2
11
1 2 1
1
x− 3 y +
x− 3 +3 ,
n = 1, 2 · · · ,
for all x, y ∈ A.
3
n
2
n
By using the same method in the above examples, we easily get that all conditions of Theorem
4.3 are satisfied. By virtue of Theorem 4.3, we have In → I in the sense of Painlevé-Kuratowski.
However, ∀ x ∈ A, there exists y = −x ∈ A with y = x, such that Fn (x, y) + Fn (y, x) ∈
−int C, which implies that Fn (·, ·) is not C-strictly monotone on A×A. Thus, Theorems 4.1–4.2
and the corresponding results in [11] are not applicable.
Convergences of the Solution Sets for Perturbed Vector Equilibrium Problems without Monotonicity
857
Moreover, we can also see that Theorem 3.2 in [15], Lemma 2.1 and Theorem 2.1 in [16]
(and even Theorems 4.1–4.2 of [27]) are not applicable, where the strong assumption of C-strict
monotonicity is required.
5
Conclusion
In this paper, under new assumptions, which are weaker than C-strict monotonicity, we establish
sufficient conditions for the Painlevé-Kuratowski Convergence of the weak efficient solution sets
and efficient solution sets for the perturbed vector equilibrium problem with a sequence of
mappings where f -solution set is a general set-valued one. These results extend and improve
the corresponding ones obtained in [11, 15, 16]. Some examples are also given to illustrate the
cases.
References
[1] Aubin, J.P., Ekeland, I. Applied Nonlinear Analysis. John Wiley and Sons, New York, 1984
[2] Aubin, J.P., Frankowska, H. Set-Valued Analysis. Birkhanser, Boston, 1990
[3] Anh, L.Q., Khanh, P.Q. On the stability of the solution sets of general multivalued vector quasiequilibrium
problems. Journal of Optimization Theorey and Applications, 135: 271–284 (2007)
[4] Berge, C. Topological Spaces. Oliver and Boyd, London, 1963
[5] Chen, C.R., Li, S.J., Teo, K.L. Solution semicontinuity of parametric generalized vector equilibrium problems. Joural of Global Optimization, 45: 309–318 (2009)
[6] Chen, C.R., Li, S.J. On the solution continuity of parametric generalized systems. Pacific Journal of
Optimization, 6: 141–151 (2010)
[7] Cheng, Y.H., Zhu, D.L. Global stability results for the weak vector variational inequality. Joural of Global
Optimization, 32: 543–550 (2005)
[8] Chen, G.Y., Huang, X.X., Yang, X.Q. Vector Optimization: Set-valued and Variational Analysis. Springer,
Berlin, 2005
[9] Durea, M. On the existence and stability of approximate solutions of perturbed vector equilibrium problems.
Journal of Mathematical Analysis and Applications, 333: 1165–1179 (2007)
[10] Fang, Z.M., Li, S.J., Teo, K.L. Painlevé-Kuratowski Convergence for the solution sets of set-valued weak
vector variational inequalities. Journal of Inequalities and Applications, Volume 2008, Article ID 435719,
14 pages (2008)
[11] Fang, Z.M., Li, S.J. Painlevé-Kuratowski Convergence of the solution sets to perturbed generalized systems.
Acta Mathematicae Applicatae Sinica, English Series, 2: 361–370 (2012)
[12] Giannessi, F. Vector variational inequalities and vector equilibria. Mathematical Theories. Kluwer,
Dordrecht, 2000
[13] Giannessi, F., Maugeri, A., Pardalos, P.M. Equilibrium Problems: Nonsmooth Optimization and Variational Inequality Models. Kluwer Academi Publishers, Dordrecht, 2001
[14] Gong, X.H. Efficiency and Henig efficiency for vector equilibrium problems. Journal of Optimization
Theorey and Applications, 108: 139–154 (2001)
[15] Gong, X.H. Connectedness of the solution sets and scalarization for vector equilibrium problems. Journal
of Optimization Theorey and Applications, 133: 151–161 (2007)
[16] Gong, X.H., Yao, J.C. Lower semicontinuity of the set of efficient solutions for generalized systems. Journal
of Optimization Theorey and Applications, 138: 197–205 (2008)
[17] Gong, X.H. Continuity of the solution set to parametric weak vector equilibrium problems. Journal of
Optimization Theorey and Applications, 139: 35–46 (2008)
[18] Gong, X.H., Yao, J.C. Connectedness of the set of efficient solutions for generalized systems. Journal of
Optimization Theorey and Applications, 138: 189–196 (2008)
[19] Huang, N.J., Li, J., Thompson, H.B. Stability for parametric implicit vector equilibrium problems. Mathematical and Computer Modelling, 43: 1267–1274 (2006)
[20] Huang, X.X. Stability in vector-valued and set-valued optimization. Mathematical Methods of Operations
Research, 52: 185–195 (2000)
[21] Lignola, M.B., Morgan, J. Generalized variational inequalities with pseudomonotone opterators under
perturbations. Journal of Optimization Theorey and Applications, 101: 213–220 (1999)
[22] Lucchetti, R.E., Miglierina, E. Stability for convex vector optimization problems. Optimization, 53:
517–528 (2004)
858
Z.Y. PENG, X.M. YANG
[23] Li, S.J., Fang, Z.M. Lower semicontinuity of the solution mappings to a parametric generalized Ky Fan
inequality. Journal of Optimization Theorey and Applications, 147: 507–515 (2010)
[24] Li, S.J., Chen, G.Y., Teo, K.L. On the stability of generalized vector quasivariational inequality problems.
Journal of Optimization Theorey and Applications, 113: 283–295 (2002)
[25] Li, Z.H. Benson Proper Efficiency in the Vector Optimization of Set-Valued Maps. Journal of Optimization
Theorey and Applications, 98: 623–649 (1998)
[26] Oppezzi, P., Rossi, A.M. A convergence for vector-valued functions. Optimization, 57: 435–448 (2008)
[27] Peng, Z.Y., Yang, X.M. Semicontinuity of the solution mappings to weak generalized parametric Ky Fan
inequality problems with trifunctions. Optimization, DOI:10.1080/02331934.2012.660693, 2012
[28] Rockafellar, R.T., Wets, R.J. Variational analysis. Springer-Verlag, Berlin, 1998