AAA Abstract and Applied Analysis 1687-0409 1085-3375 Hindawi Publishing Corporation 176363 10.1155/2013/176363 176363 Research Article Strong Duality and Optimality Conditions for Generalized Equilibrium Problems 0000-0003-4825-407X Fang D. H. 1, 2 Bao J. F. 3 Yao Jen-Chih 1 College of Mathematics and Statistics Jishou University Jishou 416000 China jsu.edu.cn 2 Department of Applied Mathematics National Sun Yat-sen University Kaohsiung 80424 Taiwan nsysu.edu.tw 3 School of Mathematics Physics and Information Science Zhejiang Ocean University Zhoushan 316004 China 2013 3 10 2013 2013 29 08 2013 04 09 2013 2013 Copyright © 2013 D. H. Fang and J. F. Bao. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We consider a generalized equilibrium problem involving DC functions. By using the properties of the epigraph of the conjugate functions, some sufficient and/or necessary conditions for the weak and strong duality results and optimality conditions for generalized equilibrium problems are provided.

1. Introduction

Consider the following generalized equilibrium problem: (1)(GEP)FindxCsuchthatf(x,y)+ψ(y)ψ(x)foreachyC, where X is a locally convex Hausdorff topological space, C is a nonempty convex subset of X, f:X×X¯{+}, and ψ:X¯ are proper functions satisfying the following.

f(x,x)=0 for all xC.

fx(·)f(x,·) is proper convex for all xC.

ψg-h, where g,h:X¯ are two proper convex functions.

Here and throughout the whole paper, following [1, page 39], we adapt the convention that (+)+(-)=(+)-(+)=+.

As mentioned in , equilibrium problems theory provides us with a unified, natural, innovative, and general framework to study a wide class of problems arising in finance, economics, network analysis, transportation, elasticity, and optimization. This theory has witnessed an explosive growth in theoretical advances and applications across all disciplines of pure and applied sciences. Equilibrium problems have been studied extensively, and many problems such as optimization problems, Nash equilibria problems, complementarity problems, fixed point problems, variational inequality problems, and convex vector optimization problems can be recast into the form (GEP); see, for example,  and the references therein.

Duality for equilibrium problems was first studied in . The schemes proposed in that paper are extensions of a classical duality theory for variational inequalities. In spirit of convex optimization, duality results and optimality conditions have been obtained for equilibrium problems by Martínez-Legaz and Sosa  when ψ=0 and by Jacinto and Scheimberg  when ψ is convex, which extended the classical convex duality results. Recently, the authors in  considered the generalized equilibrium problems in the case where ψ is a DC function. Under the assumptions that C is closed and functions f(·,x), g, h are lower semicontinuous (lsc in brief), they gave some weak and strong duality results and optimality conditions for (GEP) via a closedness qualification condition.

Inspired by the works mentioned above, we continue to study the generalized equilibrium problems. Our main aim in the present paper is to give some new regularity conditions which characterize the weak duality, the strong duality, and optimality conditions for (GEP). In general, we do not impose any topological assumption on C or on fx(·), g, and h; that is, C is not necessarily closed, and fx(·), g, and h are not necessarily lsc. Most of results obtained in the present paper are proper extensions of the early results (e.g., [5, 1416]). In particular, even in the special case when the closedness of C and lower semicontinuity of fx(·), g, and h are satisfied, our results provide improved versions of [5, Theorems 4.2, 4.3 and 4.4].

The paper is organized as follows. Section 2 contains some necessary notations and preliminary results. In Section 3, we develop general duality and optimality results for a DC optimization problem. The weak and strong duality results and optimality conditions for generalized equilibrium problems are given in Section 4.

2. Notations and Preliminaries

The notation used in the present paper is standard (cf. ). In particular, we assume throughout the whole paper that X is a real locally convex Hausdorff topological vector space, and X* denotes the dual space, endowed with the weak*-topology w*(X*,X). By x*,x we will denote the value of the functional x*X* at xX; that is, x*,x=x*(x). Let Z be a set in X. The closure of Z is denoted by clZ. If WX*, then clW denotes the weak*-closure of W. For the whole paper, we endow X*× with the product topology of w*(X*,X) and the usual Euclidean topology.

The indicator function δZ of a nonempty set Z is defined by δZ(x)=0 if xZ and δZ(x)=+ if xZ. Let f:X¯ be a proper function. The conjugate function and the epigraph of f are denoted by f* and epif, respectively; they are defined by (2)f*(x*)supxdomf{x*,x-f(x)}foreachx*X*,epif{(x,r)X×:f(x)r}, where the effective domain domf{xX:f(x)<+}. It is well known and easy to verify that epif* is weak*-closed. The lsc hull of f, denoted by clf, is defined by (3)epi(clf)=cl(epif). Then, by [1, Theorem 2.3.1], f*=(clf)* and by [1, Theorem 2.3.4], if clf is proper and convex, then f**=clf. Let xX. The subdifferential of f at x is defined by (4)f(x){x*X*:f(x)+x*,y-xf(y)foreachyXf(x)+x*,y-x} if xdomf, and f(x) otherwise. Clearly, the following equivalence holds: (5)x0isaminimizeroffiff0f(x0). By definition, the Young-Fenchel inequality below holds: (6)f(x)+f*(x*)x,x*foreachpair(x,x*)X×X*. Moreover, by [1, Theorem 2.4.2(iii)], (7)f(x)+f*(x*)=x*,xiffx*f(x) (the equality in (7) is usually referred to as Young's equality). If g, h are proper, then (8)epig*+epih*epi(g+h)*,ghg*h*epig*epih*,g(a)+h(a)(g+h)(a)foreachadomgdomh. Furthermore, for each pX* and a, (9)epi(h+p+a)*=epih*+(p,-a). Moreover, if g is convex and lsc on domh, then, by [17, Lemma 2.3], (10)epi(h-g)*=x*domg*(epih*-(x*,g*(x*))).

The following lemma is known in [1, 6] (cf. [6, Lemma 2.1] for (11) and (12) and [1, Theorem 2.8.7] for (13)).

Lemma 1.

Let g,h:X¯ be proper convex functions satisfying domgdomh.

If g, h are lsc, then (11)epi(g+h)*=cl(epig*+epih*).

If either g or h is continuous at some point of domgdomh, then (12)epi(g+h)*=epig*+epih*,(13)(g+h)(x)=g(x)+h(x)foreachxdomgdomh.

3. Duality and Optimality Conditions for DC Optimization Problem

Let pX*. Consider the following DC optimization problem: (14)(Pp)infyC{F(y)+G(y)-H(y)-p,y}, where C is a convex subset of X (not necessarily closed) and F,G,H:X¯ are proper convex functions (not necessaily lsc). Following , we define the dual problem of (Pp) by (15)(Dp)infx*domH*supu*domF*,v*domG*Lp(x*,u*,v*), where Lp:domH*×domF*×domG*¯ is defined by (16)Lp(x*,u*,v*)H*(x*)-F*(p+u*)-G*(v*)-δC*(x*-u*-v*). Let v(Pp) and v(Dp) denote the optimal values of problems (Pp) and (Dp), respectively. Especially, in the case when p=0, we write (P), (D) and L(·,·,·) for (P0), (D0), and L0(·,·,·), respectively. One of the main aims in this section is devoted to the study of the weak duality and the strong duality between (P) and (D), which are defined as follows.

Definition 2.

We say that

the weak duality holds (between (P) and (D)) if v(D)v(P);

the strong duality holds (between (P) and (D)) if v(P)=v(D) and for each x*domF*, there exists (u*,v*)domF*×domG* such that L(x*,u*,v*)v(D);

the stable weak duality (resp., the stable strong duality) holds if the weak duality (resp., the strong duality) holds between (Pp) and (Dp) for each pX*.

If H is lsc, then by [5, Theorem 3.2(i)], the weak duality holds. However, the weak duality does not necessarily hold in general as will be shown in the following example.

Example 3.

Let X and C[0,+). Let F,G,H:¯ be defined by F0, Gδ[0,+), and (17)H(y){0y<0,1y=0,+y>0. Then, F, G, and H are proper convex functions and v(P)=-1. Clearly, F*=δ{0}, G*=δC*=δ(-,0], and H*=δ[0,+). Hence, v(D)=0. This implies that v(D)>v(P). Consequently, the weak duality does not hold.

To consider the weak duality, the strong duality, and optimality conditions for problem (P), we introduced the following conditions. For simplicity, we denote (18)Kx*domH*((x*,H*(x*))epiF*+epiG*+epiδC*-(x*,H*(x*))), where we adapt the convention tSt=X.

Definition 4.

The family {F,G,H,δC} is said to satisfy

the weak closure condition at 0 ((WCC)0) if (19)K({0}×)epi(F+G-H+δC)*({0}×);

the closure condition at 0 ((CC)0) if (20)K({0}×)=epi(F+G-H+δC)*({0}×);

the weak closure condition ((WCC)) if (21)Kepi(F+G-H+δC)*;

the closure condition ((CC)) if (22)K=epi(F+G-H+δC)*;

the Moreau-Rockafellar formula (MRF) at xdom(F+G-H)C if (23)(F+G-H+δC)(x)x*H(x)(F(x)+G(x)+NC(x)-x*);

(MRF) if it satisfies (MRF) at each point in dom(F+G-H)C.

Remark 5.

If H is lsc, then by (10), we have that (24)K=epi(F-H)*+epiG*+epiδC* and, by (8), (21) holds; that is, the (WCC) holds.

The following proposition describes the relationship between the (CC) (resp., the (WCC)) and the (CC)0 (resp., the (WCC)0).

Proposition 6.

The family {F,G,H,δC} satisfies the (CC) (resp., the (WCC)) if and only if for each pX*, {F-p,G,H,δC} satisfies the (CC)0 (resp., the (WCC)0).

Proof.

Let pX*, and let K(p) be the set defined by (25)K(p)x*domH*((x*,H*(x*))epi(F-p)*+epiG*+epiδC*-(x*,H*(x*))). Then, by (9), the following equality is clear: (26)K(p)=K+(-p,0). Hence, we have that (27)K(p)({0}×)=K({p}×)+(-p,0). Moreover, using (9), we conclude that (28)epi(F-p+G-H+δC)*({0}×)=epi(F+G-H+δC)*({p}×)+(-p,0). Thus, the conclusion holds by definitions and the proof is complete.

Under the assumption that (29)F,G,HarelscandCisclosed, the authors in  introduced the following closure condition (30)epiF*+epiG*+epiδC*isweak*-closed       to consider the strong duality and optimality conditions for DC optimization problem (14). The following proposition describes the relationships among the (CC), the (MRF), and (30).

Proposition 7.

The following implication holds: (31)(CC)(MRF). Furthermore, if (29) holds, then the following implications hold: (32)(30)(CC)(MRF).

Proof.

Suppose that the (CC) holds. Let x0A, and let p(F+G-H+δC)(x0). Then, by (7), (33)(p,p,x0-(F+G-H+δC)(x0))epi(F+G-H+δC)*=K, thanks to the (CC). Hence, for each x*H(x0), (34)(p,p,x0-(F+G-H+δC)(x0))epiF*+epiG*+epiδC*-(x*,H*(x*)). Let x*H(x0). Then, there exist (u*,r1)epiF*, (v*,r2)epiG*, and (w*,r3)epiδC* such that (35)u*+v*+w*-x*=p,r1+r2+r3-H*(x*)=p,x0-(F+G-H+δC)(x0). Below we show that u*F(x0), v*G(x0), and w*NC(x0). To do this, note by the definition that (36)F*(u*)r1,G*(v*)r2,δC*(w*)r3. Moreover, since x*H(x0), it follows from (7) that (37)H(x0)+H*(x*)=x*,x0. Hence, by (35)–(37) and the Young-Fenchel inequality (6), we have that (38)0F*(u*)+F(x0)-u*,x0F*(u*)+F(x0)-x*+p-v*-w*,x0F*(u*)-r1+(v*,x0-G(x0)-r2)+(w*,x0-δC(x0)-r3)+(H(x0)+H*(x*)-x*,x0)(F*(u*)-r1)+(G*(v*)-r2)+(δC*(w*)-r3)0. Thus, (39)F*(u*)+F(x0)-u*,x0=0. This implies that u*F(x0) by (7). Using the same argument, we have that v*G(x0) and w*NC(x0). Hence, pF(x0)+G(x0)+NC(x0)-x*, and (40)px*H(x0)(F(x0)+G(x0)+NC(x0)-x*) since x*H(x0) is arbitrary. Therefore, the (MRF) holds.

Furthermore, suppose that (29) holds. To show that (32), we only need to show the implication (30)(CC) holds. To do this, we assume that (30) holds. Since H is lsc, it follows from (10) that (41)epi(F+G+δC-H)*=x*domH*(epi(F+G+δC)*-(x*,H*(x*)))  . Note that F, G are lsc and C is closed; by Lemma 1(i), one has that (42)epi(F+G+δC)*=cl(epiF*+epiG*+epiδC*). This together with (30) and (41) implies that the (CC) holds. The proof is complete.

To study the weak duality and the strong duality, we need the following lemma.

Lemma 8.

Let r. Then, the following assertions hold:

(0,r)epi(F+G-H+δC)* if and only if v(P)-r.

(0,r)K if and only if v(D)-r and for each x*domH*, there exist u*domF* and v*domG* such that (43)L(x*,u*,v*)-r.

Proof.

(i) By the definition of the conjugate function, one has (44)v(P)=-(F+G-H+δC)*(0). Hence, the result is clear.

(ii) Let (0,r)K and let x*domH*. Then (45)(0,r)epiF*+epiG*+epiδC*-(x*,H*(x*)). Thus, there exist (u*,r1)epiF*, (v*,r2)epiG*, and (w*,r3)epiδC* such that (46)u*+v*+w*-x*=0,r1+r2+r3-H*(x*)=r. Since (47)F*(u*)r1,G*(v*)r2,δC*(w*)r3, it follows from (46) that (48)L(x*,u*,v*)=H*(x*)-F*(u*)-G*(v*)-δC*(x*-u*-v*)H*(x*)-r1-r2-r3=-r. This together with the definition of v(D) implies that v(D)-r and u*, v* satisfy (43).

Conversely, suppose that v(D)-r and for each x*domH*, there exist u*domF* and v*domG* satisfying (43). Let x*domH*. Then, there exist u*domF* and v*domG* such that (43) holds. Then (49)δC*(x*-u*-v*)r+H*(x*)-F*(u*)-G*(v*). This means that (50)(x*-u*-v*,r+H*(x*)-F*(u*)-G*(v*))epiδC*. Therefore, (51)(0,r)=(u*,F*(u*))+(v*,G*(v*))-(x*,H*(x*))+(x*-u*-v*,r+H*(x*)-F*(u*)-G*(v*))epiF*+epiG+epiδC*-(x*,H*(x*)). Noting that x*domH* is arbitrary, we have that (52)(0,r)x*domH*((x*,H*(x*))epiF*+epiG+epiδC*-(x*,H*(x*)))=K. Thus, we complete the proof.

Our first theorem of this section shows that the (WCC) is a sufficient and necessary condition for the weak duality to hold.

Theorem 9.

(i) The weak duality holds if and only if the family {F,G,H,δC} satisfies the (WCC)0.

(ii) The stable weak duality holds if and only if the family {F,G,H,δC} satisfies the (WCC).

Proof.

As assertion (ii) is a global version of assertion (i). Hence, by Proposition 6, it suffices to prove assertion (i). Suppose that the weak duality holds. Let (0,r)K. Then, by Lemma 8(ii), we have v(D)-r and hence v(P)-r, which implies that (0,r)(F+G-H+δC)*, thanks to Lemma 8(i). Hence, (19) holds; that is, the (WCC)0 holds.

Conversely, suppose that the family {F,G,H,δC} satisfies the (WCC)0. To show that v(D)v(P), suppose on the contrary that v(P)<v(D). Then, there exists r such that v(P)<-rv(D). Thus, by the definition of v(D), we have that for each x*domH*, there exist u*domF* and v*domG* such that (43) holds. Hence, (0,r)K by Lemma 8(ii), and (0,r)epi(F+G-H+δC)* by the (WCC)0. This together with Lemma 8(i) implies that v(P)-r, which contradicts to v(P)<-r. Consequently, we have v(D)v(P) and complete the proof.

Theorem 10.

(i) The strong duality holds if and only if the family {F,G,H,δC} satisfies the (CC)0.

(ii) The stable strong duality holds if and only if the family {F,G,H,δC} satisfies the (CC).

Proof.

As before, it is sufficient to prove assertion (i). Suppose that the strong duality holds. Let x*domH*. Then, v(P)=v(D) and there exist u*domF* and v*domG* such that L(x*,u*,v*)v(D). By Theorem 9(i), (WCC)0 holds, and so, we only need to verify the following inclusion: (53)epi(F+G-H+δC)*({0}×)K({0}×). To do this, let (0,r)epi(F+G-H+δC)*. Then, by Lemma 8(i), we have v(P)-r. Hence, v(D)=v(P)-r and x*domH*, u*domF*, v*domG* satisfying (43). This together with Lemma 8(ii) implies that (0,r)K as x*domH* is arbitrary. Hence, (53) holds and so does the (CC)0.

Conversely, suppose that the (CC)0 holds. Then, the family {f,g,δC;ft,gt:tT} satisfies (WCC)0, and so v(D)v(P) by Theorem 9(i). Thus, to prove the strong duality, it suffices to show that v(D)v(P) and that for each x*domH* there exist u*domF* and v*domG* satisfying L(x*,u*,v*)v(D). Note that the conclusion holds trivially if v(P)=-. Below we only consider the case when -rv(P). By Lemma 8(i), (0,r)epi(F+G-H+δC)*, and so (0,r)K, thanks to the (CC)0. Then, by Lemma 8(ii) and the definition of v(D), we have that v(D)-r and for each x*domH* there exist u*domF* and v*domG* satisfying L(x*,u*,v*)-r. Hence, the strong duality holds. The proof is complete.

Theorem 11.

Let x0 be a solution of (P). Suppose that the family {F,G,H,δC} satisfies the (MRF) at x0. Then (54)H(x0)F(x0)+G(x0)+NC(x0). Furthermore, if H(x0), then for each x*H(x0), there exist u*domF* and v*domG* such that (55)v(P)=L(x*,u*,v*).

Proof.

Since x0 is a solution of (P), it follows that (56)0(F+G-H+δC)(x0). Then, by the (MRF) at x0, one has that (57)0x*H(x0)(F(x0)+G(x0)+NC(x0)-x*), which is equivalent to (54) holds.

Furthermore, assume that x*H(x0). Then by (54), there exist u*F(x0), v*G(x0), and w*NC(x0) such that (58)x*=u*+v*+w*. Since x*H(x0), u*F(x0), v*G(x0), and w*NC(x0), it follows from (7) that (59)x*,x0=H*(x*)+H(x0),u*,x0=F*(u*)+F(x0),v*,x0=G*(v*)+G(x0),w*,x0=δC*(u*). Hence, (60)L(x*,u*,v*)=F(x0)+G(x0)-H(x0)=v(P), which completes the proof.

Remark 12.

In the case when (29) holds, Dinh et al. established the weak duality and strong duality in [5, Theorem 3.2] and the optimality condition in [5, Theorem 3.1] under the assumption that (30) holds. Clearly, by Proposition 7, Theorems 9 and 10 extend and improve [5, Theorem 3.2] and Theorem 11 extends and improves [5, Theorem 3.1].

4. Optimality Conditions and Dualities for Equilibrium Problem

Recall the optimization problem (GEP) is defined as in Section 1. Let xC and consider the DC optimization problem (61)(𝒫x)p(x)infyC{f(x,y)+ψ(y)}. Then, by the definitions of (GEP) and (𝒫x), a point x0C is a solution of (GEP) if and only if x0 is a solution of (𝒫x0) (cf. [5, Lemma 3.1]). Moreover, by the definition, we can find that p(x)ψ(x) for each xC, and, x0 is a solution of (GEP) if and only if p(x0)=ψ(x0). Hence, the problem of finding solutions of (GEP) can be reduced to the one of finding solutions of the following optimization problem: (62)(𝒫)maxxCinfyC{fx(y)+ψ(y)-ψ(x)}. Following , we defined the dual problem of (𝒫) by (63)(𝒟)maxxCinfx*domh*maxu*domfx*,v*domg*{L(x,x*,u*,v*)22222222222222222222222222222222-ψ(x)}, where L:C×X*×X*×X*¯ is defined by (64)L(x,x*,u*,v*)h*(x*)-fx*(u*)-g*(v*)-δC*(x*-u*-v*).

Let v(𝒫) and v(𝒟) denote the optimal values of problems (𝒫) and (𝒟), respectively. Unlike , the weak duality (i.e., v(𝒟)v(𝒫)) does not necessarily holds in general. Recall from Definition 4 that for each xC, the family {fx,g,h,δC} satisfies the (WCC)0 if (65)x*domh*(epifx*+epig*+epiδC*-(x*,h*(x*)))({0}×)epi(fx+g-h+δC)*({0}×) and it satisfies the (CC)0 if (66)x*domh*(epifx*+epig*+epiδC*-(x*,h*(x*)))({0}×)=epi(fx+g-h+δC)*({0}×). Then, we have the following theorem.

Theorem 13.

(i) Suppose that for each xC, the family {fx,g,h,δC} satisfies the (WCC)0. Then, v(𝒟)v(𝒫).

(ii) Suppose that for each xC, the family {fx,g,h,δC} satisfies the (CC)0. Then, v(𝒟)=v(𝒫).

Proof.

(i) Since for each xC, the family {fx,g,h,δC} satisfies the (WCC)0, it follows from Theorem 9(i) that (67)p(x)infx*domh*maxu*domfx*,v*domg*L(x,x*,u*,v*). Hence, by the definitions of v(𝒫) and v(𝒟), we see that v(𝒟)v(𝒫).

(ii) Since for each xC, the family {fx,g,h,δC} satisfies the (CC)0, it follows from Theorem 10(i) that (68)p(x)=infx*domh*maxu*domfx*,v*domg*L(x,x*,u*,v*). Thus, the result is seen to hold.

The following theorems establish the relationships between the solutions of (GEP) and those of (𝒟). First, we recall that a point x0C is said to be a solution of (GEP) if (69)f(x0,y)+ψ(y)ψ(x0)foreachyC, and it is said to be a solution of the dual problem (𝒟) if for each x*domh*, there exist u*,v*X* such that (70)L(x0,x*,u*,v*)=ψ(x0)=g(x0)-h(x0). partially, if for each x*h(x0), there exist u*,v*X* such that (70) holds, then x0 is said to be a weak solution of problem (𝒟).

Remark 14.

(a) Obviously, x0C is a solution of (GEP) if and only if x0 is a solution of (𝒫x0).

(b) Let x0C. If h is lsc at x0, then for each x*domh*, there exist u*domfx0* and v*domg* such that (71)L(x0,x*,u*,v*)g(x0)-h(x0). Consequently, x0 is a solution of the dual problem (𝒟) if and only if for each x*domh*, there exist u*,v*X* such that (72)L(x0,x*,u*,v*)g(x0)-h(x0). In fact, let x*domh*, u*domfx0*, and v*domg*. Then by the Young-Fenchel inequality (6), it is easy to see that (73)L(x0,x*,u*,v*)h*(x*)-x*,x0+g(x0)+fx0(x0)infx*domh*{h*(x*)-x*,x0}+g(x0)+fx0(x0)=-(clh)(x0)+g(x0)=g(x0)-h(x0). Hence, (71) holds. However, (71) does not necessarily hold in general as will be showen in the following example.

Example 15.

Let X and C(-,0]. Let f:×¯ and g,h:¯ be defined by f0, gδ[0,+), and (74)h(x){0x<0,1x=0,+x>0. Let x0=0. Then, g(x0)-h(x0)=-1, and h is not lsc at x0. Clearly, fx0*=δ{0}, g*=δ(-,0], and h*=δC*=δ[0,+). Then, domfx0*={0}, domg*=(-,0], and domh*=domδC*=[0,+). Let x*=0. Then, for each u*domfx0* and v*domg*, one has that (75)L(x0,x*,u*,v*)=-g*(v*)-δC*(-v*)=0; this implies that (71) does not hold.

By Theorem 11, we obtain the the following theorem straightforwardly, which was established in [5, Theorem 4.2] under the assumptions that (29) holds and (76)epifx*+epig*+epiδC*isweak*-closed. Thus, by Proposition 7, our Theorem 16 improves the corresponding result in [5, Theorem 4.2].

Theorem 16.

Let x0C such that H(x0). Suppose that the family {fx0,g,h,δC} satisfies the (MRF) at x0. If x0 is a solution of (GEP), then x0 is a weak solution of (𝒟).

Theorem 17.

Let x0C. Suppose that h is lsc and that the family {fx0,g,h,δC} satisfies the (CC)0. Then, x0 is a solution of (GEP) if and only if x0 is a solution of the dual problem (𝒟).

Proof.

Suppose that x0 is a solution of (GEP). Then, x0 is a solution of (𝒫x0). Hence, (77)v(𝒫x0)=fx0(x0)+g(x0)-h(x0)=g(x0)-h(x0). Since the family {fx0,g,h,δC} satisfies the (CC)0, it follows from Theorem 10 that (78)v(𝒫x0)=infx*domh*maxu*domfx0*,v*domg*L(x0,x*,u*,v*), which implies that for each x*domh*, there exist u*domfx0* and v*domg* such that (79)L(x0,x*,u*,v*)v(Px0). Moreover, since h is lsc at x0, it follows from Remark 14(b) that (80)L(x0,x*,u*,v*)g(x0)-h(x0)=v(Px0). Hence, (70) holds. This means that x0 is a solution of the dual problem (𝒟).

Conversely, suppose that x0 is a solution of the dual problem (𝒟). Let x*domh*. Then, there exist u*domfx0* and v*domg* such that (70) holds. Note by the Young-Fenchel inequality (6) that for each xC, (81)L(x0,x*,u*,v*)h*(x*)-x*,x+g(x)+fx0(x). Then by (70), one has that for each xC, (82)g(x0)-h(x0)h*(x*)-x*,x+g(x)+fx0(x). Since the above inequality holds for each x*domh*, it follows that (83)g(x0)-h(x0)infx*domh*{h*(x*)-x*,y}+g(x)+fx0(x)=-(clh)(x)+g(x)+fx0(x). Consequently, by the lower semicontinuity of h, one has that for each xC, (84)g(x0)-h(x0)fx0(x)+g(x)-h(x), and hence g(x0)-h(x0)p(x0). This implies that g(x0)-h(x0)=p(x0) since g(x0)-h(x0)p(x0) holds automatically. Thus, x0 is a solution of (GEP).

Remark 18.

Theorem 17 was established in [5, Theorems 4.3 and 4.4] under the assumptions that (29) and (76) hold. Thus, our Theorem 16 extends the corresponding result in [5, Theorems 4.3 and 4.4] to suit the case when g, h are not lsc and C is not closed.

Below we will give a upper estimate for the Fréchet subdifferential of the function p defined in (61). We first recall from  or [19, page 90] that the Fréchet subdifferential of ϕ at a point x0domϕ with |ϕ(x0)|< is defined by (85)^ϕ(x0){x*X*:liminfxx0ϕ(x)-ϕ(x0)-x*,x-x0x-x00}, where ϕ:X{±} is an extended real-valued function.

Theorem 19.

Let x0C be a solution of (GEP). Suppose that the family {g,δC} satisfies the (MRF) at x0; that is, (86)(g+δC)(x0)=g(x0)+NC(x0). Then for each γ>0, one has that (87)p^(x0)x*h(x0)(g(x0)+NC(x0)-x*)+γ𝔹*, where 𝔹* denotes the unit ball in X*.

Proof.

Let x0C. If H(x0)=, then (87) holds trivially. Below let u*^p(x0) and x*h(x0). Let γ>0. Then, by the definition of of Fréchet subdifferential of p at x0, there exists η>0 such that (88)p(x)-p(x0)-u*,x-x0+γx-x00foreachxx0+η𝔹, where 𝔹 denotes the unit ball in X. Since x0 is a solution of (GEP), it follows that p(x0)=g(x0)-h(x0). Note that p(x)g(x)-h(x) holds for each xC. Then (89)0g(x)-h(x)-g(x0)+h(x0)-u*,x-x0+γx-x0g(x)-g(x0)-u*+x*,x-x0+γx-x0, where the last inequality holds because x*h(x0). Define ϕ:X¯ by (90)ϕ(x)g(x)-g(x0)-u*+x*,x-x0+γx-x0. Then, ϕ is a proper convex function on X, and x0 is a solution to the following convex programming: (91)Minϕ(x),s.t.xC(x0+η𝔹). Hence, 0(ϕ+δC(x0+η𝔹))(x0), and (92)0(g+δC(x0+η𝔹))(x0)-(u*+x*)+γ𝔹*. Note that x0 is an interior point of the set x0+η𝔹; then the function δx0+η𝔹 is continuous at x0. Hence, (93)(g+δC(x0+η𝔹))(x0)=(g+δC)(x0). This together with (86) and (92) implies that (94)0g(x0)+NC(x0)-(u*+x*)+γ𝔹*; that is, (95)u*g(x0)+NC(x0)-x*+γ𝔹*, Which implies that (87) holds by the arbitrariness of x*. The proof is complete.

Conflict of Interests

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

Acknowledgments

D. H. Fang was supported in part by the National Natural Science Foundation of China (Grant no. 11101186) and supported in part by the Scientific Research Fund of Hunan Provincial Education Department (Grant no. 13B095).

Zălinescu C. Convex Analysis in General Vector Spaces 2002 River Edge, NJ, USA World Scientific Publishing 10.1142/9789812777096 MR1921556 Noor M. A. Noor K. I. On equilibrium problems Applied Mathematics E-Notes 2004 4 125 132 MR2077792 ZBL1064.49009 Bigi G. Castellani M. Kassay G. A dual view of equilibrium problems Journal of Mathematical Analysis and Applications 2008 342 1 17 26 10.1016/j.jmaa.2007.11.034 MR2440776 ZBL1155.90021 Blum E. Oettli W. From optimization and variational inequalities to equilibrium problems The Mathematics Student 1994 63 1–4 123 145 MR1292380 ZBL0888.49007 Dinh N. Strodiot J. J. Nguyen V. H. Duality and optimality conditions for generalized equilibrium problems involving DC functions Journal of Global Optimization 2010 48 2 183 208 10.1007/s10898-009-9486-z MR2721761 ZBL1228.90078 Fang D. H. Li C. Ng K. F. Constraint qualifications for extended Farkas's lemmas and Lagrangian dualities in convex infinite programming SIAM Journal on Optimization 2009 20 3 1311 1332 10.1137/080739124 MR2546346 ZBL1206.90198 Iusem A. N. Sosa W. Iterative algorithms for equilibrium problems Optimization 2003 52 3 301 316 10.1080/0233193031000120039 MR1995678 ZBL1176.90640 Muu L. D. Nguyen V. H. Quy N. V. On Nash-Cournot oligopolistic market equilibrium models with concave cost functions Journal of Global Optimization 2008 41 3 351 364 10.1007/s10898-007-9243-0 MR2424472 ZBL1146.91029 Muu L. D. Oettli W. Convergence of an adaptive penalty scheme for finding constrained equilibria Nonlinear Analysis. Theory, Methods & Applications 1992 18 12 1159 1166 10.1016/0362-546X(92)90159-C MR1171603 ZBL0773.90092 Nguyen T. T. V. Strodiot J. J. Nguyen V. H. A bundle method for solving equilibrium problems Mathematical Programming 2009 116 1-2 529 552 10.1007/s10107-007-0112-x MR2421293 ZBL1155.49006 Konnov I. V. Schaible S. Duality for equilibrium problems under generalized monotonicity Journal of Optimization Theory and Applications 2000 104 2 395 408 10.1023/A:1004665830923 MR1752324 ZBL1016.90066 Martínez-Legaz J. E. Sosa W. Duality for equilibrium problems Journal of Global Optimization 2006 35 2 311 319 10.1007/s10898-005-3840-6 MR2242017 ZBL1106.90074 Jacinto F. M. O. Scheimberg S. Duality for generalized equilibrium problem Optimization 2008 57 6 795 805 10.1080/02331930701761458 MR2473942 ZBL1152.90648 Altangerel L. Boţ R. I. Wanka G. On the construction of gap functions for variational inequalities via conjugate duality Asia-Pacific Journal of Operational Research 2007 24 3 353 371 10.1142/S0217595907001309 MR2335551 ZBL1141.49303 Altangerel L. Boţ R. I. Wanka G. On gap functions for equilibrium problems via Fenchel duality Pacific Journal of Optimization 2006 2 3 667 678 MR2265761 ZBL1103.49016 Mosco U. Dual variational inequalities Journal of Mathematical Analysis and Applications 1972 40 202 206 MR0313913 10.1016/0022-247X(72)90043-1 ZBL0277.49011 Fang D. H. Li C. Yang X. Q. Stable and total Fenchel duality for DC optimization problems in locally convex spaces SIAM Journal on Optimization 2011 21 3 730 760 10.1137/100789749 MR2837549 ZBL1236.90140 Mordukhovich B. S. Nam N. M. Yen N. D. Fréchet subdifferential calculus and optimality conditions in nondifferentiable programming Optimization 2006 55 5-6 685 708 10.1080/02331930600816395 MR2274933 ZBL1121.49017 Mordukhovich B. S. Variational Analysis and Generalized Differentiation 2006 330 Berlin, Germany Springer MR2191744