See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269353629

E-Benson proper efficiency in vector optimization

Article in Optimization · September 2013

DOI: 10.1080/02331934.2013.798321

CITATIONS READS

33 221

2 authors, including:

Xin Min Yang

Chongqing Normal University
96 PUBLICATIONS 2,145 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

stochastic programming View project

All content following this page was uploaded by Xin Min Yang on 29 August 2015.

The user has requested enhancement of the downloaded file.

This article was downloaded by: [Shanghai University]
On: 04 March 2014, At: 18:57
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Optimization: A Journal of
Mathematical Programming and
Operations Research
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gopt20

-Benson proper efficiency in vector

optimization
a a
Ke Quan Zhao & Xin Min Yang
a
Department of Mathematics , Chongqing Normal University ,
Chongqing , China
Published online: 06 Jun 2013.

To cite this article: Ke Quan Zhao & Xin Min Yang (2013): -Benson proper efficiency in vector
optimization, Optimization: A Journal of Mathematical Programming and Operations Research, DOI:
10.1080/02331934.2013.798321

To link to this article: http://dx.doi.org/10.1080/02331934.2013.798321

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 Optimization, 2013
http://dx.doi.org/10.1080/02331934.2013.798321

E-Benson proper efficiency in vector optimization

Ke Quan Zhao∗ and Xin Min Yang

Department of Mathematics, Chongqing Normal University, Chongqing, China

(Received 7 August 2012; final version received 11 April 2013)

Starting from the innovative ideas of Chicco et al. (Vector optimization problems
via improvement sets. J. Optim. Theory Appl. 2011;150:516–529), in this paper,
the concepts of improvement set and E-efficiency are introduced in a real lo-
Downloaded by [Shanghai University] at 18:57 04 March 2014

cally convex Hausdorff topological vector space. Furthermore, some properties

of the improvement sets are given and a kind of proper efficiency, named as
E-Benson proper efficiency, which unifies some proper efficiency and approx-
imate proper efficiency, is proposed via the improvement sets in vector opti-
mization. Moreover, the concept of E-subconvexlikeness of set-valued maps is
introduced via the improvement sets and an alternative theorem is proved. In
the end, some scalarization theorems and Lagrange multiplier theorems of E-
Benson proper efficiency are established for a vector optimization problem with
set-valued maps.
Keywords: vector optimization; E-Benson proper efficiency; E-subconvexlikeness;
scalarization; Lagrange multipliers
AMS Subject Classifications: 90C26; 90C29; 90C30

1. Introduction
It is well known that the theory and methods of vector optimization have been playing an
important role in almost all optimal decision problems. During the recent years, there have
been a lot of studies in the area of vector optimization, see e.g. [1–3] and the references
therein. The concept of efficient solution has played a useful role in analyzing vector
optimization problems. But, sometimes the efficient solution set includes some points of a
certain anomalous type and has some undesirable consequences. Thus, some concepts of
proper efficiency have been proposed to restrict the efficient solution set. Various kinds of
classic proper efficiency concepts include Geoffrion proper efficiency,[4] Benson proper
efficiency,[5] Borwein proper efficiency,[6] Henig proper efficiency,[7] super efficiency,[8],
etc. In particular, under some suitable assumptions of generalized convexity, some authors
studied some characterizations of the Benson proper efficiency in terms of scalarization,
Lagrange multiplier theorems, saddle-point criteria and duality, see e.g. [9–12].
The concepts of approximate solution have been playing an important role when there
are no exact solutions in optimization problems. Kutateladze initially introduced the concept
of approximate solution named as ε-efficient solution in [13]. In recent years, some concepts
of approximate efficiency and their applications were studied in vector optimization, see
e.g. [14–19] and the references therein.

*Corresponding author. Email: [email protected]

© 2013 Taylor & Francis

 2 K.Q. Zhao and X.M. Yang

Very recently, Chicco et al. proposed a new concept of approximate efficiency named as
E-efficiency based on upper comprehensive sets in [20]. E-efficiency unifies the concepts
of optimal points, approximate optimal points in scalar optimization, Pareto equilibria and
approximate equilibra. Liu introduced the concept of ε-proper efficiency by the idea of
Geoffrion proper efficiency and obtained some scalarization results in [21]. Rong and
Ma proposed the concept of ε-proper efficiency by the idea of Benson proper efficiency
and established the scalarization theorems, some ε-Lagrange multipliers theorems, some
ε-saddle-point theorems and some ε-proper duality theorems in [22]. Gao et al. proposed a
kind of approximate proper efficiency of vector optimization problems and obtained some
necessary and sufficiency conditions via some scalarizations in [23]. But, the approximate
proper efficiency has some undesirable characterizations; for example, it does not necessar-
ily imply approximate efficiency. Gutierrez et al. proposed nearly (C, ε)-subconvexlikeness
and studied scalarization and saddle points of this kind of approximate proper efficiency in
[24].
Downloaded by [Shanghai University] at 18:57 04 March 2014

Motivated by the works of [9–12,17,20,22,23], in this paper, we first present the concepts
of improvement sets and E-efficiency in a real locally convex Hausdorff topological vector
space. We obtain some properties of improvement sets, which will be important to derive the
main results, and present the concept of E-Benson proper efficiency via the improvement
sets in vector optimization. E-Benson proper efficiency unifies some exact and approximate
proper efficiency. Furthermore, based on the idea of cone-subconvexlikeness, we introduce
the concept of E-subconvexlikeness of set-valued maps and obtain an alternative theorem.
In the end, we establish some scalarization theorems and Lagrange multiplier theorems of
E-Benson proper efficiency for a vector optimization problem with set-valued maps.

2. Preliminaries
Throughout this paper, let X be a linear space, Y and Z be real locally convex Hausdorff
topological vector spaces. For a set A ⊆ Y , intA and clA denote interior and closure of A,
respectively. A cone K ⊆ Y is called pointed if K ∩ (−K ) = {0}. Let K and P be positive
closed convex pointed cones in Y and Z with non-empty interiors, respectively. For any
x, y ∈ Y
x K y ⇔ y − x ∈ K .

The generated cone of a set A ⊂ Y is cone(A) = {λa| λ ≥ 0, a ∈ A}. Y ∗ denotes the

topological dual space of Y . The positive dual cone of a subset A ⊂ Y is defined as

A+ := {μ ∈ Y ∗ | y, μ ≥ 0, ∀y ∈ A},

the quasi-interior of A+ is defined as

A+i := {μ ∈ Y ∗ | y, μ > 0, ∀y ∈ A\{0}},

where y, μ denotes the value of the linear functional μ at y.

Chicco et al. introduce the concepts of the improvement sets and E-efficiency in a finite
dimensional space in [20]. In the following, we propose some corresponding concepts
in a real locally convex Hausdorff topological vector space. Furthermore, Rn denotes n
dimensional Euclidean space and Rn+ denotes its non-negative orthant.
 Optimization 3

Definition 2.1 The upper comprehensive set of a subset A in Y is defined as

u-compr(A) = {x ∈ Y | there is a ∈ A s.t. x − a ∈ K }.

Definition 2.2 A subset E ⊂ Y is called an upper comprehensive set if

u-compr(E) = E.

Definition 2.3 If E ⊂ Y \{0} is an upper comprehensive set, then we call E an improve-

ment set of Y . We denote the family of the improvement sets in Y by using TY .

Remark 2.1 It is easy to prove that E ∈ TY if and only if 0 ∈

/ E and E + K = E. Let
Y = C(G) be the space of continuous real-valued functions on G, where G ⊂ Rn is a
bounded closed set, let K = {x ∈ C(G)|x(t) ≥ 0, ∀ t ∈ G}. Then, E = {x ∈ C(G)|x(t) >
0, ∀ t ∈ G} is an improvement set.
Downloaded by [Shanghai University] at 18:57 04 March 2014

Clearly, ∅ ∈ TY . In this paper, we suppose that intE = ∅.

Remark 2.2 Let Y = Rm , K = Rm + . Then, Definition 2.3 coincides with Definition 2.5 in
[20]. We denote the family of the improvement sets in Rm by using Tm .

Based on the ideas of Chicco et al. in [20], we propose the concept of E-efficient point
for minimum problems.

Definition 2.4 Let E ∈ TY and a non-empty set A ⊂ Y be given. We say that a ∈ A is an

E-efficient point if (a − E) ∩ A = ∅ and denote this by using a ∈ O E (A).

3. Some properties of the improvement sets

In this section, we give some properties of improvement set. These properties are important
to establish our main results.

Remark 3.1 From the definition of improvement set and Remark 2.1, the following state-
ments are obvious.

(i) K \{0} ∈ TY and E + K \{0} ∈ TY ;

(ii) If ε ∈ K \{0}, then ε + K ∈ TY ;
(iii) If E ⊂ K and E ∈ TY , then E + E ⊂ E and intE + E ⊂ E; 
(iv) If E ⊂ K , E ∈ TY and μ ∈ K +i , then μE ∈ T1 , where μE = μ, e ; and
e∈E
(v) If E ∈ TY , then intE ∈ TY .

Theorem 3.1 If E ∈ TY , then int E = E + intK = clE + intK .

Proof The proof of intE = E + intK is similar with Lemma 3.4 in [25], so is omitted.
Clearly, E + intK ⊂ clE + intK . We only need to prove E + intK ⊃ clE + intK . For any
given x ∈ clE + intK , then there exist x 1 ∈ clE and x2 ∈ intK such that x = x1 + x2 .
From x2 ∈ intK , it follows that there exists a balanced neighbourhood of zero V such that
x2 + V ⊂ intK . Hence, from x1 ∈ clE, we also have (x1 + V ) ∩ E = ∅. So, there exists
v ∈ V such that x1 + v ∈ E and
x + v + V = x1 + v + x2 + V ⊂ E + intK .
 4 K.Q. Zhao and X.M. Yang

It is clear that x + v + V is a neighbourhood of x, which implies that x ∈ int(E + intK ) ⊂

E + intK , i.e. E + intK ⊃ clE + intK . 
Flores-Bazán and Hernández proposed Assumption (B) and studied some scalarization
results in [26]. The following Remark 3.2 shows the fact that the concept of improvement
set is a special case of Assumption (B) proposed by Flores-Bazán and Hernández.
Remark 3.2
(i) First of all, we point out the fact that if E is an improvement set in Y and q ∈ intK ,
then
cl(C(−E)) + R++ q ⊂ intC(−E),
where C(−E) denotes the complement of the set −E. In fact, we only need to prove
the following implication holds:
Y \(−intE) + R++ q ⊂ Y \(−clE).
Downloaded by [Shanghai University] at 18:57 04 March 2014

For any given x ∈ Y \(−intE) + R++ q. Clearly, x ∈ Y . So, we need to prove

x∈/ −clE. On the contrary, suppose that −x ∈ clE. Since x ∈ Y \(−intE) + R++ q,
then there exist x1 ∈ Y \(−intE) and x2 ∈ R++ q such that x = x1 + x2 . From
Theorem 3.1, it follows that
−x1 = −x + x2 ∈ clE + R++ q ⊂ clE + intK = intE,
which contradicts to x1 ∈ Y \(−intE).
(ii) If a set E ⊂ Y satisfies Assumption (B), then it is not necessarily an improvement set.
In fact, let E = R2++ ∪ {(0, 1)} ∪ {(1, 0)}. We can verify that E satisfies Assumption
(B). However, it is clear that E is not an improvement set.
(iii) There exists an improvement set such that 0 ∈ / ∂ E. For instance, let E = {(x 1 , x2 ) ∈
R2 |x1 ≥ 0.1, x2 ≥ 0.1}, it is clear that E ∈ T2 and 0 ∈ / ∂ E. However as in [26]
the condition that the origin lies on the boundary is not restrictive, since it can be
translated and the translation satisfies Assumption (B).
Remark 3.3 If E is not an improvement set, then Theorem 3.1 is not necessarily true. In
fact, let Y = R2 , K = R2+ , E = {(x1 , x2 )|x1 ≤ 0, x2 ≤ 0}. Clearly, E ∈
/ T2 and
intE = {(x1 , x2 )|x1 < 0, x2 < 0} = E + intK = R2 .
Remark 3.4 If E ∈ TY , then from Remark 3.1 (v) and Theorem 3.1, E + intK ∈ TY .
Theorem 3.2 If E ∈ TY and A ⊂ Y , then clcone(A + E) = clcone(A + int E).
Proof Since E ∈ TY , then from Proposition 4.1(ii) in [10] and Theorem 3.1 we have
clcone(A + E) = clcone(A + E + K ) = clcone(A + E + intK ) = clcone(A + intE).

Theorem 3.3 Let E ∈ TY , E ⊂ K and μ ∈ K +i . Then, int(μE) = μint E.
Proof From E ∈ TY , Remark 3.1(iv) and Theorem 3.1, we have μE ∈ T1 and
int(μE) = μE + intR+ = μE + μintK = μ(E + intK ) = μintE.

 Optimization 5

4. E-Benson proper efficiency

In this section, we propose the concept of E-Benson proper efficiency via improvement
sets by using the idea of the classic Benson proper efficiency.
In Definition 2.4, let Y = R, ε > 0 and E = ε + R+ . From Remark 3.1(ii), we have
E ∈ T1 . Thus, E-efficiency for minimum problems implies that for any y ∈ A, y > a − ε.
This means that a is a strict approximate optimal point of the set A (see [27]). Furthermore,
if E ∈ TY , from Remark 3.1(i), we know that E + K \{0} ∈ TY . Based on these facts, in
this paper, we consider the following definition of E-efficiency for minimum problems.

Definition 4.1 Let E ∈ TY and a non-empty set A ⊂ Y be given. We say that a ∈ A is an

E-efficient point of A if (a − E − K \{0}) ∩ A = ∅ and we denote this by a ∈ O E+K \{0} (A).

Definition 4.2 Let E ∈ TY and a non-empty set A ⊂ Y be given. We say that a ∈ A is an

E-Benson proper efficient point of A if
Downloaded by [Shanghai University] at 18:57 04 March 2014

clcone(A + E − a) ∩ (−K ) = {0}. (4.1)

We denote this by a ∈ O BES (A).

Remark 4.1 By Definitions 4.1 and 4.2, it is obvious that

O BES (A) ⊂ O E+K \{0} (A).

Indeed, for any given a ∈ O BES (A), from (4.1), it follows that

(A + E − a) ∩ (−K ) ⊂ {0}.

Thus,
(A + E − a) ∩ (−K \{0}) = ∅.
This shows that a ∈ O E+K \{0} (A). 

Remark 4.2 The following examples illustrate the converse relation is not necessarily true
in Remark 4.1.

Example 4.1 Let Y = R2 , K = R2+ , A = {(x1 , x2 )|x1 = 0, x2 < 0} ∪ {(x1 , x2 )|0 ≤ x1 ≤

1, 0 ≤ x2 ≤ 1} and E = {(x1 , x2 )|x1 ≥ 0.1, x2 ≥ 0.1}. Clearly, E ∈ T2 . From Definition
4.1, it is clear that (0, 0) ∈ O E+K \{0} (A). But, since

clcone(A + E − (0, 0)) ∩ (−K ) = {(x 1 , x2 )|x1 = 0, x2 ≤ 0} = {(0, 0)}

/ O BES (A).
and from Definition 4.2, it follows that (0, 0) ∈

Example 4.2 Let Y = R2 , K = R2+ , A = {(x1 , x2 )|x1 = 0, x2 > 0} ∪ {(x1 , x2 )|0 ≤

x1 ≤ 1, −1 ≤ x2 ≤ 0} and E = {(x1 , x2 )|x1 ≥ 0.1, x2 ≥ 0.1}. Clearly, E ∈ T2 . From
Definition 4.1 and Definition 4.2, we have (0, 0) ∈ O E+K \{0} (A) and (0, 0) ∈ O BES (A).

Remark 4.3 If E = K \{0}, then E-Benson proper efficiency coincides with the classic
Benson proper efficiency. In fact, by Remark 3.1(i), E ∈ TY . From Proposition 4.1(ii) in
[10], we obtain that
 6 K.Q. Zhao and X.M. Yang

clcone(A + E − a) = clcone(A + K \{0} − a)

⊂ clcone(A + K − a)
= clcone(A + intK − a)
⊂ clcone(A + K \{0} − a).
This implies that clcone(A + E − a) = clcone(A + K − a). From (4.1), it follows that
clcone(A + K − a) ∩ (−K ) = {0}, i.e. a is a Benson proper efficient point of A.

Remark 4.4 Let ε ∈ K \{0} and E = ε + K . From Remark 3.1(ii), E ∈ TY and E-Benson
proper efficiency coincides with ε-proper efficiency introduced by Rong et al. in [22].

Remark 4.5 E-Benson proper efficiency unifies some exact and approximate proper effi-
ciency, besides the classic Benson proper efficiency and the ε-proper efficiency.
Downloaded by [Shanghai University] at 18:57 04 March 2014

Example 4.3 Let Y = R2 , K = R2+ , E = {(x1 , x2 )|x1 ≥ 1, x2 ≥ 0} ∪ {(x1 , x2 )|x1 ≥

0, x2 ≥ 1} and A = {(x1 , x2 )|0 ≤ x1 ≤ 1, −1 ≤ x2 ≤ 0} ∪ {(x1 , x2 )|x1 + x2 = 0, x1 ≤ 0}.
Clearly, E = K and E = ε + K ,where, ε ∈ K \{0}. Furthermore, (0, 0) ∈ O BES (A).

Remark 4.6 E-Benson proper efficiency is different from (C, ε)-proper efficiency pro-
posed by Gutierrez et al. in [24]. In fact, from Definition 2.1 in [24] and Definition 4.2, we
only need to verify that C(ε) is different from E.

Example 4.4

(i) Let Y = R2 , K = R2+ and E = R2 \R2− . Clearly, E ∈ T2 , but for any C ⊂ K \{0},
ε ≥ 0, E = C(ε).
1 3
(ii) Let Y = R2 , K = R2+ and C = {(x1 , x2 )| x1 ≤ x2 ≤ x1 , x1 ≥ 0}\{(0, 0)}.
2 2
Clearly, C ⊂ K \{0}. But ∀ε ≥ 0, C(ε) is not an improvement set.

5. Vector optimization with set-valued maps

In this paper, we consider the following vector optimization problem with set-valued maps:

(VP) min F(x),

s.t. x ∈ D = {x ∈ S|G(x) ∩ (−P) = ∅}
, where S ⊂ X , F : S ⇒ Y and G : S ⇒ Z are set-valued maps with non-empty value.
We assume that the feasible set D of (VP) is non-empty. Denote

 
F(x), μ = { y, μ |y ∈ F(x)}, F(A) = F(x), F(A), μ = F(x), μ .
x∈A x∈A

where μ ∈ Y ∗ and A ⊂ X . Let L(Z , Y ) be the space of continuous linear operator from Z
to Y and L + = L + (Z , Y ) = {T ∈ L(Z , Y )|T (P) ⊂ K }. Denote by (F, G) the set-valued
map from S to Y × Z defined by (F, G)(x) = F(x) × G(x). If μ ∈ Y ∗ , T ∈ L(Z , Y ), we
also define μF : S ⇒ R and F + T G : S ⇒ Y by

(μF)(x) = F(x), μ and (F + T G)(x) = F(x) + T (G(x)),

 Optimization 7

respectively. Moreover, we say that (VP) satisfies the generalized slater constraint qualifi-
cation (see [9]) if there exists x̂ ∈ S such that G(x̂) ∩ (−intP) = ∅.

Definition 5.1 [20] A point x0 ∈ D is called an E-efficient solution of (VP) if there exists
y0 ∈ F(x0 ) such that
(y0 − E − K \{0}) ∩ F(D) = ∅.

Definition 5.2 A point x0 ∈ D is called an E-Benson proper efficient solution of (VP) if

F(x0 ) ∩ O BES (F(D)) = ∅.

Definition 5.3 A pair (x0 , y0 ) is called an E-Benson proper efficient point of (VP) if

x0 ∈ D and y0 ∈ F(x0 ) ∩ O BES (F(D)).

Downloaded by [Shanghai University] at 18:57 04 March 2014

Example 5.1 Let X = Y = Z = R2 , K = P = R2+ , E = {(x1 , x2 )|x1 ≥ 1, x2 ≥

0} ∪ {(x1 , x2 )|x1 ≥ 0, x2 ≥ 1} and S = {(x1 , x2 )|x1 = 0, −1 ≤ x2 ≤ 0}. F(x) : R2 → R2
and F(x) = x. G(x) : R2 → R2 and G(x) = x. Clearly, D = S and (0, 0) is an E-Benson
proper efficient point of (VP).

Remark 5.1 From Remarks 4.3 and 4.4, we can see that Definition 5.2 and Definition
5.3 unify Definition 2.1 in [9] and Definition 2.1 and Definition 2.2 in [22]. Furthermore,
E-Benson proper efficiency of (VP) implies E-efficiency of (VP), and the converse is not
necessarily true. This can be seen through Remarks 4.1 and 4.2.

6. E-subconvexlikeness
In [28], Li and Chen introduced the concept of convexlikeness for set-valued maps and sub-
convexlikeness for set-valued maps and established an alternative theorem for generalized
inequality–equality systems of set-valued maps. In this section, we propose the concept
of E-subconvexlikeness for set-valued maps via the improvement sets and establish an
alternative theorem for E-subconvexlike set-valued maps.

Definition 6.1 Let F : S ⇒ Y be a set-valued map and E ∈ TY . F is said to be E-

subconvexlike on S if F(S) + intE is a convex set.

Remark 6.1 If E such that intK ⊂ E ⊂ K \{0}, then E-subconvexlikeness coincides with
K -subconvexlikeness. In fact, we have F(S) + intE = F(S) + intK .

Remark 6.2 E-subconvexlikeness does not necessarily imply K -subconvexlikeness. The

following example illustrates this point.

Example 6.1 Let S = {(x1 , x2 )|x1 ≤ 0, x1 + x2 = 0} ∪ {(x1 , x2 )|x1 ≥ 1, x1 + x2 = 0},

Y = R2 , K = R2+ and E = R2+ ∩ {(x1 , x2 )|x1 + x2 ≥ 1}. F : R2 → R2 and F(x) = x.
Clearly, F(S) + intE is a convex set, but F(S) + intK is not a convex set.

Theorem 6.1 If E ∈ TY and the map F is E-subconvexlike on S, then exactly one of

the following statements is true:
 8 K.Q. Zhao and X.M. Yang

(i) ∃ x ∈ S, F(x) ∩ (−intE) = ∅;

(ii) ∃ μ ∈ K + \{0Y ∗ }, y + e, μ ≥ 0, ∀y ∈ F(S), ∀e ∈ E.

Proof If both (i) and (ii) hold, from Theorem 3.1, then there exist x ∈ S, y ∈ F(x) and
e ∈ E such that 0 ≤ y + e, μ < 0, which is a contradiction. Consequently, assume that
(i) holds, then (ii) does not hold. Now, assume that (i) does not hold. Then, again from
Theorem 3.1, we can prove that

0∈
/ F(S) + intE = F(S) + E + intK . (6.1)

From the E-subconvexlikeness of F, the set F(S)+ intE is convex. Hence, by the separation
theorem for convex sets and from (6.1), there exists μ ∈ Y ∗ \{0} such that

y + e + εr, μ ≥ 0, ∀y ∈ F(S), ∀e ∈ E, ∀r ∈ intK , ∀ε > 0. (6.2)

Letting ε → +∞ in (6.2), we have r, μ ≥ 0, ∀r ∈ intK . Consequently, r, μ ≥ 0, ∀r ∈

Downloaded by [Shanghai University] at 18:57 04 March 2014

K . Hence, μ ∈ K + \{0Y ∗ }. Letting ε ↓ 0 in (6.2), we have y +e, μ ≥ 0, ∀y ∈ F(S), ∀e ∈

E. Hence, (ii) holds.

Theorem 6.2 Let E ∈ TY , E ⊂ K and k0 ∈ intK . If (F, G) is (E × P)-subconvexlike

on S, then for μ ∈ K +i , (μF, G) is (μE × P)-subconvexlike on S.

Proof The proof of the theorem is easy to obtain from Theorem 3.3, so is omitted.

7. Scalarization
In this section, we consider the following scalar optimization problem of (VP):

(VP)μ min F(x), μ , μ ∈ Y ∗ \{0Y ∗ }.

x∈D

The following definition unifies Definition 4.1 in [9] and Definition 3.1 in [22].

Definition 7.1 Let E ∈ TY . A point x0 ∈ D is called an E-optimal solution of (VP)μ if

there exists y0 ∈ F(x0 ) such that

( y0 , μ − μ(E + K \{0})) ∩ F(D), μ = ∅. (7.1)

The point pair (x0 , y0 ) is called an E-optimal point of (VP)μ .

If μ ∈ K +i , then (7.1) is equivalent to y−y0 +e, μ ≥ 0, ∀x ∈ D, ∀y ∈ F(x), ∀e ∈ E.

Remark 7.1 If ε ∈ K \{0}, E = ε + K and μ ∈ K +i , then E ∈ TY and Definition 7.1

reduces to y, μ ≥ y0 , μ − ε, μ , ∀x ∈ D, ∀y ∈ F(x).

Remark 7.2 Even if μ ∈ K +i , μ(E + K \{0}) is not necessarily an improvement set in R.

Example 7.1 Let Y = R2 , K = R2+ and  E  = {(x, y)|x + y ≥ 0}\{(0, 0)}. Clearly,
2 1
E ∈ T2 and E + K \{0} ∈ T2 . Let μ = , ∈ K +i and z = (−1, 2) ∈ E + K \{0}.
3 3
Thus z, μ = 0 ∈ μ(E + K \{0}). This means that μ(E + K \{0}) is not an improvement
set in R.
 Optimization 9

Remark 7.3 If μ ∈ K +i , E ∈ TY and E ⊂ K , then μ(E + K \{0}) is an improvement set

in R. Consequently, Definition 7.1 implies that y0 , μ ∈ O μ(E+K \{0}) ( F(D), μ ).

Lem m a 7.1 [6] Let N , K ⊂ Y be two closed and convex cones such that N ∩ K = {0}.
If K is pointed and locally compact, then (−N + ) ∩ (K +i ) = ∅.

Theorem 7.1 Let μ ∈ K +i and E ∈ TY . If (x0 , y0 ) is an E-optimal point of (VP)μ , then

(x0 , y0 ) is an E-Benson proper efficient point of (VP).

Proof Let d ∈ clcone(F(D) + E − y0 ) ∩ (−K ). Since E ∈ TY , we have

d ∈ clcone(F(D) + E + K − y0 ) ∩ (−K ).

Hence, there exist {yn } ⊂ F(D), {en } ⊂ E, {rn } ⊂ K and {λn } ⊂ R+ such that

d = lim λn (yn + en + rn − y0 ). (7.2)

Downloaded by [Shanghai University] at 18:57 04 March 2014

n→∞

Since μ ∈ K +i and from (7.2), we have

d, μ = lim λn ( yn , μ + en , μ + rn , μ − y0 , μ ). (7.3)
n→∞

Since (x0 , y0 ) is an E-optimal point of (VP)μ , we can obtain that yn , μ + en , μ −

y0 , μ ≥ 0. On the other hand, rn , μ ≥ 0. Thus, from (7.3), it follows that d, μ ≥ 0.
Moreover, since μ ∈ K +i and d ∈ (−K ), it is clear that d, μ ≤ 0. This yields d, μ = 0.
Furthermore, we have d = 0. Hence, y0 ∈ O BES (F(D)). From y0 ∈ F(x0 ), it follows that
y0 ∈ F(x0 ) ∩ O BES (F(D)). Therefore, (x0 , y0 ) is an E-Benson proper efficient point of
(VP).

Theorem 7.2 Let E ∈ TY and F be E-subconvexlike on D. If (x 0 , y0 ) is an E-Benson

proper efficient point of (VP), then there exists μ ∈ K +i such that (x0 , y0 ) is an E-optimal
point of (VP)μ .

Proof Since (x0 , y0 ) is an E-Benson proper efficient point of (VP), we have x 0 ∈ D, y0 ∈

F(x0 ) and clcone(F(D) + E − y0 ) ∩ (−K ) = {0}. From Theorem 3.2, we obtain that

clcone(F(D) + intE − y0 ) ∩ (−K ) = {0}. (7.4)

From the E-subconvexlikeness of F, we know that F(D) + intE is convex. Thus,

clcone(F(D) + intE − y0 ) is a closed convex cone. Hence, from (7.4) and Lemma 7.1,
there exists μ ∈ K +i such that

z, μ ≥ 0, ∀z ∈ clcone(F(D) + intE − y0 ). (7.5)

Furthermore, again from Theorem 3.2, we have

F(D) + E − y0 ⊂ clcone(F(D) + E − y0 ) = clcone(F(D) + intE − y0 ). (7.6)

From (7.5) and (7.6), it follows that y + e − y0 , μ ≥ 0, ∀x ∈ D, ∀y ∈ F(x), ∀e ∈ E.

This implies that (x0 , y0 ) is an E-optimal point of (VP)μ .
The set of all E-Benson proper efficient point of (VP) is denoted by E B P E(V P) and
the set of all E-optimal point of (VP)μ is denoted by E A P(V P)μ . 
 10 K.Q. Zhao and X.M. Yang

Corollary 7.1 Let E ∈ Tm and F be E-subconvexlike on D. Then,


E B P E(V P) = E A P(V P)μ .
μ∈K +i

Proof It is a direct result of Theorems 7.1 and 7.2.

Remark 7.4 If E = K \{0}, then from Remarks 4.3 and 6.1, it follows that Corollary 7.1
reduces to Corollary 4.1 in [9].

In Theorems 7.1 and 7.2, let E = ε + K , where ε ∈ K \{0}. We can obtain that the following
corollary.

Corollary 7.2 Let ε ∈ K \{0} and F be K -subconvexlike on D. Then, (x 0 , y0 ) is an

ε-properly efficient point of (VP) iff there exists μ ∈ K +i such that
Downloaded by [Shanghai University] at 18:57 04 March 2014

y, μ ≥ y0 , μ − ε, μ , ∀x ∈ D, ∀y ∈ F(x).

8. Lagrange multipliers
In this section, we present Lagrange multiplier theorems of E-Benson proper efficiency of
problem (VP).

Theorem 8.1 Let E ∈ TY , E ⊂ K , F be E-subconvexlike on D and (F, G) be

(E × P)-subconvexlike on S. Furthermore, let (VP) satisfy the generalized slater constraint
qualification. If (x0 , y0 ) is an E-Benson proper efficient point of (VP) and 0 ∈ G(x 0 ), then
there exists T ∈ L + (Z , Y ) such that −T (G(x0 ) ∩ (−P)) ⊂ (intK ∪ {0})\intE and (x 0 , y0 )
is an E-Benson proper efficient point of the following unconstrained vector optimization
problem with set-valued maps:
(UVP) min F(x) + T (G(x)),
s.t. x∈S
Proof Since (x0 , y0 ) is an E-Benson proper efficient point of (VP), by Theorem 7.2, there
exists μ ∈ K +i such that
y − y0 + e, μ ≥ 0, ∀x ∈ D, ∀y ∈ F(x), ∀e ∈ E. (8.1)
Define H : S ⇒ R × Z by
H (x) = F(x) − y0 , μ × G(x) = (μF, G)(x) − ( y0 , μ , 0 Z ).
Since (F, G) is (E × P)-subconvexlike on S from Theorem 6.2, it follows that H is
(μE × P)-subconvexlike on S. Moreover, (8.1) implies that the system
x ∈ S, H (x) ∩ (−int(μE × P)) = ∅
has no solution. Hence, from Theorem 6.1, there exists (λ, ϕ) ∈ R+ × P\{(0, 0 Z )} such
that
λ y − y0 + e, μ + z + z  , ϕ ≥ 0, ∀x ∈ S, ∀y ∈ F(x), ∀z ∈ G(x), ∀e ∈ E, ∀z  ∈ P.
(8.2)
 Optimization 11

In particular, let z  = 0 in (8.2), we can obtain that

λ y − y0 + e, μ + z, ϕ ≥ 0, ∀x ∈ S, ∀y ∈ F(x), ∀z ∈ G(x), ∀e ∈ E. (8.3)

From the fact that the map G satisfies the generalized slater constraint qualification and (8.3),
we can easily show that λ > 0. Take an appropriate k0 ∈ intK satisfying λ k0 , μ = 1. We
define a map T : Z → Y as follows

T (z) = z, ϕ k0 , z ∈ Z. (8.4)

Clearly, T ∈ L + (Z , Y ). Setting x = x0 , y = y0 and z = z 0 ∈ G(x0 ) ∩ (−P) in (8.3), we

can obtain that
−λ e, μ ≤ z 0 , ϕ ≤ 0. (8.5)
From the right inequality of (8.5), we have that
Downloaded by [Shanghai University] at 18:57 04 March 2014

−T (z 0 ) = − z 0 , ϕ k0 ∈ intK ∪ {0}.

From the left inequality of (8.5), we have that −T (z 0 ) ∈

/ intE. Otherwise, from Theorem
3.1, there exists e ∈ E such that −T (z 0 )−e ∈ intK . Consequently, T (z 0 )+e, μ < 0, i.e.
z 0 , ϕ + λ e, μ < 0, which contradicts to the left inequality of (8.5). Notice that z 0 is
arbitrary in the set G(x0 ) ∩ (−P), we obtain that

−T (G(x0 ) ∩ (−P)) ⊂ (intK ∪ {0})\intE.

Furthermore, from T ∈ L + (Z , Y ) and 0 ∈ G(x0 ), it follows that 0 ∈ T (G(x0 )). Thus,

y0 ∈ F(x0 ) ⊂ F(x0 ) + T (G(x0 )). Hence, from (8.3) and (8.4), it follows that
λ y + T (z), μ = λ y, μ + λ z, ϕ k0 , μ
= λ y, μ + z, ϕ
≥ λ y0 − e, μ , ∀x ∈ S, ∀y ∈ F(x), ∀z ∈ G(x), ∀e ∈ E.
Dividing the above inequality by λ > 0 leads to

y + T (z), μ − y0 , μ + e, μ ≥ 0, ∀x ∈ S, ∀y ∈ F(x), ∀z ∈ G(x), ∀e ∈ E.

Hence, (x0 , y0 ) is an E-optimal point of (UVP)μ given by

(UVP)μ min F(x) + T (G(x)), μ .

x∈S

Again by μ ∈ K +i and Theorem 7.1, (x0 , y0 ) is an E-Benson proper efficient point of

(UVP).

Theorem 8.2 Let E ∈ TY , x0 ∈ D and y0 ∈ F(x0 ). If there exists T ∈ L + (Z , Y )

such that 0 ∈ T (G(x0 )) and (x0 , y0 ) is an E-Benson proper efficient point of (UVP), then
(x0 , y0 ) is an E-Benson proper efficient point of (VP).

Proof From the assumptions, we have y0 ∈ F(x0 ) ⊂ F(x0 ) + T (G(x0 )) and

  

clcone (F(x) + T (G(x))) + E − y0 ∩ (−K ) = {0}. (8.6)
x∈S
 12 K.Q. Zhao and X.M. Yang

Since
x ∈ D ⇒ G(x) ∩ (−P) = ∅
⇒ ∃z ∈ G(x) s.t. z ∈ (−P)
⇒ −T (z) ∈ K
⇒ K ⊂ T (z) + K ⊂ T (G(x)) + K ,
and E ∈ TY , we have

F(D) + E − y0 = (F(x) + E − y0 )
x∈D
⊂ (F(x) + E + T (G(x)) + K − y0 )

x∈D
⊂ (F(x) + T (G(x))) + E − y0 .
x∈S
Hence,
Downloaded by [Shanghai University] at 18:57 04 March 2014

 

0 ∈ clcone (F(D) + E − y0 ) ⊂ clcone (F(x) + T (G(x))) + E − y0 .
x∈S

Combining with (8.6), this yields clcone(F(D) + E − y0 ) ∩ (−K ) = {0}. Since x0 ∈ D,

y0 ∈ F(x0 ) ⊂ F(D), we obtain that y0 ∈ F(x0 ) ∩ O BES (F(D)). Hence, (x0 , y0 ) is an
E-Benson proper efficient point of (VP).

Remark 8.1 If E = K \{0}, then from Remarks 4.3 and 6.1, it follows that Theorems 8.1
and 8.2 reduce to Theorems 5.1 and 5.2 in [9], respectively.

In the end, we present the following example to illustrate Theorems 8.1 and 8.2.

Example 8.1 Let X = Y = Z = R2 , E = {(y1 , y2 )|y1 ≥ 0.1, y2 ≥ 0.1}, S =

[−0.1, 0.1] × {0}. The set-valued maps F : S ⇒ Y and G : S ⇒ Z are, respectively,
defined as follows:
F(x1 , x2 ) = {|x1 |} × [0.1 − |x1 |, 0.2 − |x1 |], (x1 , x2 ) ∈ [−0.1, 0.1] × {0};

{(z 1 , z 2 |(z 1 , z 2 ) = λ(−0.1, x1 ), λ ∈ [0, 1]} if (x1 , x2 ) ∈ [−0.1, 0] × {0},
G(x1 , x2 ) =
[0, x1 ] × {0} if (x1 , x2 ) ∈ (0, 0.1] × {0}.
Let x0 = (0.1, 0) and y0 = (0.1, 0.1). Clearly, (x0 , y0 ) is an E-Benson proper efficient point
of (VP). All conditions of Theorem 8.1 are satisfied. Therefore, there exists T (z 1 , z 2 ) =
(−z 2 , 0) such that −T (G(x0 ) ∩ (−P)) ⊂ (intK ∪ {0})\intE and (x 0 , y0 ) is an E-Benson
proper efficient point of (UVP). The above example can also illustrate Theorem 8.2.

9. Concluding remarks
We introduce the concepts of improvement set and E-efficiency for minimum problems in
real locally convex Hausdorff topological vector space. We also give some properties of
improvement set and propose the concept of E-Benson proper efficiency via improvement
set. The concept is based on the ideas of classic Benson proper efficiency. We present the
concept of E-subconvexlikeness of set-valued maps and obtain an alternative theorem.
 Optimization 13

We also establish scalarization theorems and Lagrange multiplier theorems of E-Benson

proper efficiency under the assumption of E-subconvexlikeness. It will be future study to
weaken the assumption of E-subconvexlikeness by using the idea of generalized cone-
subconvexlikeness.

Acknowledgements
This work is partially supported by the National Science Foundation of China (Grants 11271391,
11126348, 11171363), the Natural Science Foundation Project of Chongqing (Grant 2011BA0030
and CSTS2012jjA00002) and the Education Committee Research Foundation of Chongqing (Grant
KJ110625).

References
Downloaded by [Shanghai University] at 18:57 04 March 2014

[1] Jahn J. Vector optimization. Theory, applications, and extensions. Berlin: Springer; 2004.
[2] Chen GY, Huang XX, Yang XQ. Vector optimization. Vol. 541, Lecture Notes in Economics and
Mathematical Sciences. Berlin: Springer; 2005.
[3] Ehrgott M. Multicriteria optimization. Berlin: Springer; 2005.
[4] Geffrion AM. Proper efficiency and the theory of vector maximization. J. Math. Anal. Appl.
1968;22:618–630.
[5] Benson B. An improved definition of proper efficiency for vector maximization with respect to
cones. J. Math. Anal. Appl. 1979;71:232–241.
[6] Borwein JM. Proper efficient points for maximizations with respect to Cones. SIAM J. Control
Optim. 1977;15:57–63.
[7] Henig MI. Proper efficiency with respect to cones. J. Optim. Theory Appl. 1982;36:387–407.
[8] Borwein JM, Zhuang D. Super efficiency in vector optimization. Trans. Am. Math. Soc.
1993;338:105–122.
[9] Li ZF. Benson proper efficiency in the vector optimization of set-valued maps. J. Optim. Theory
Appl. 1998;98:623–649.
[10] Chen GY, Rong WD. Characterizations of the Benson proper efficiency for nonconvex vector
optimization. J. Optim. Theory Appl. 1998;98:365–384.
[11] Yang XM, Yang XQ, Chen GY. Theorems of the alternative and optimization with set-valued
maps. J. Optim. Theory Appl. 2000;107:627–640.
[12] Yang XM, Li D, Wang SY. Near-subconvexlikeness in vector optimization with set-valued
functions. J. Optim. Theory Appl. 2001;110:413–427.
[13] Kutateladze S. Convex ε-programming. Soviet Math. Dokl. 1979;20:391–393.
[14] White DJ. ε-efficiency. J. Optim. Theory Appl. 1986;49:319–337.
[15] Dutta J, Vetrivel V. On approximate minima in vector optimization. Numer. Funct. Anal. Optim.
2001;22:845–859.
[16] Gutirrez C, Jimnez B. A unified approach and optimality conditions for approximate solutions
of vector optimization problems. SIAM J. Optim. 2006;17:688–710.
[17] Rong WD, Wu YN. ε-Weak minimal solutions of vector optimization problems with set-valued
maps. J. Optim. Theory Appl. 2000;106:569–579.
[18] Gao Y, Hou SH, Yang XM. Existence and optimality conditions for approximate solutions to
vector optimization problems. J. Optim. Theory Appl. 2012;152:97–120.
[19] Ghaznavi-ghosoni BA, Khorram E, Soleimani-damaneh M. Scalarization for characterization of
approximate strong/weak/proper efficiency in multiobjective optimization. Optimization. 2012.
doi: 10.1080/02331934.2012.668190.
[20] Chicco M, Mignanego F, Pusillo L, Tijs S. Vector optimization problems via improvement sets.
J. Optim. Theory Appl. 2011;150:516–529.
 14 K.Q. Zhao and X.M. Yang

[21] Liu JC. ε-Properly efficient solutions to nondifferentiable multiobjective programming problems.
Appl. Math. Lett. 1999;12:109–113.
[22] Rong WD, Ma Y. ε-Properly efficient solutions of vector optimization problems with set-valued
maps. OR Trans. 2000;4:21–32.
[23] Gao Y, Yang XM. Optimality conditions for approximate solutions of vector optimization
problems. J. Ind. Manag. Optim. 2011;7:483–496.
[24] Gutierrez C, Huerga L, Novo V. Scalarization and saddle points of approximate proper
solutions in nearly subconvexlike vector optimization problems. J. Math. Anal. Appl. 2012;389:
1046–1058.
[25] Zhao KQ, Yang XM. A unified stability result with perturbations in vector optimization. Optim.
Lett. 2012. doi: 10.1007/s11590-012-0533-1.
[26] Flores-Bazán F, Hernández E. A unified vector optimization problem: complete scalarizations
and applications. Optimization. 2011;60:1399–1419.
[27] Loridan P. ε-solutions in vector minimization problems. J. Optim. Theory Appl. 1984;43:
265–276.
Downloaded by [Shanghai University] at 18:57 04 March 2014

[28] Li ZF, Chen GY. Lagrangian multipliers, saddle points, and duality in vector optimization of
set-valued maps. J. Math. Anal. Appl. 1997;215:297–316.

View publication stats