We present a projection algorithm for finding a solution of a variational inclusion problem in a real Hilbert space. Furthermore, we prove that the proposed iterative algorithm converges strongly to a solution of the variational inclusion problem which also solves some variational inequality.
1. Introduction
Let H be a real Hilbert space. Let B:H→H be a single-valued nonlinear mapping and R:H→2H be a set-valued mapping. Now we concern the following variational inclusion, which is to find a point x∈H such thatθ∈B(x)+R(x),
where θ is the zero vector in H. The set of solutions of problem (1.1) is denoted by I(B,R). If H=Rm, then problem (1.1) becomes the generalized equation introduced by Robinson [1]. If B=0, then problem (1.1) becomes the inclusion problem introduced by Rockafellar [2]. It is known that (1.1) provides a convenient framework for the unified study of optimal solutions in many optimization-related areas including mathematical programming, complementarity, variational inequalities, optimal control, mathematical economics, equilibria, game theory, and so forth. Also various types of variational inclusions problems have been extended and generalized. Recently, Zhang et al. [3] introduced a new iterative scheme for finding a common element of the set of solutions to the problem (1.1) and the set of fixed points of nonexpansive mappings in Hilbert spaces. Peng et al. [4] introduced another iterative scheme by the viscosity approximate method for finding a common element of the set of solutions of a variational inclusion with set-valued maximal monotone mapping and inverse strongly monotone mappings, the set of solutions of an equilibrium problem and the set of fixed points of a nonexpansive mapping. For some related works, see [5–28] and the references therein.
Inspired and motivated by the works in the literature, in this paper, we present a projection algorithm for finding a solution of a variational inclusion problem in a real Hilbert space. Furthermore, we prove that the proposed iterative algorithm converges strongly to a solution of the variational inclusion problem which also solves some variational inequality.
2. Preliminaries
Let H be a real Hilbert space with inner product 〈·,·〉 and norm ∥·∥. Let C be a nonempty closed convex subset of H. Recall that a mapping B:C→C is said to be α-inverse strongly monotone if there exists a constant α>0 such that 〈Bx-By,x-y〉≥α∥Bx-By∥2,for all x,y∈C. A mapping A is strongly positive on H if there exists a constant μ>0 such that 〈Ax,x〉≥μ∥x∥2 for all x∈H.
For any x∈H, there exists a unique nearest point in C, denoted by PC(x), such that‖x-PC(x)‖≤‖x-y‖,forally∈C.
Such a PC is called the metric projection of H onto C. We know that PC is nonexpansive. Further, for x∈H and x*∈C,x*=PC(x)⟺〈x-x*,x*-y〉≥0forally∈C.
A set-valued mapping T:H→2H is called monotone if, for all x,y∈H, f∈Tx and g∈Ty imply 〈x-y,f-g〉≥0. A monotone mapping T:H→2H is maximal if its graph G(T) is not properly contained in the graph of any other monotone mapping. It is known that a monotone mapping T is maximal if and only if, for (x,f)∈H×H, 〈x-y,f-g〉≥0 for every (y,g)∈G(T) implies f∈Tx.
Let the set-valued mapping R:H→2H be maximal monotone. We define the resolvent operator JR,λ associated with R and λ as follows:JR,λ=(I+λR)-1(x),x∈H,
where λ is a positive number. It is worth mentioning that the resolvent operator JR,λ is single-valued, nonexpansive, and 1-inverse strongly monotone, and that a solution of problem (1.1) is a fixed point of the operator JR,λ(I-λB) for all λ>0, see for instance [29].
Lemma 2.1 (see [30]).
Let R:H→2H be a maximal monotone mapping and B:H→H be a Lipschitz-continuous mapping. Then the mapping (R+B):H→2H is maximal monotone.
Lemma 2.2 (see [8]).
Let {xn} and {yn} be bounded sequences in a Banach space X and let {βn} be a sequence in [0,1] with 0<liminfn→∞βn≤limsupn→∞βn<1. Suppose xn+1=(1-βn)yn+βnxn for all integers n≥0 and limsupn→∞(∥yn+1-yn∥-∥xn+1-xn∥)≤0. Then, limn→∞∥yn-xn∥=0.
Lemma 2.3 (see [31]).
Assume {an} is a sequence of nonnegative real numbers such that an+1≤(1-γn)an+δn, where {γn} is a sequence in (0,1) and {δn} is a sequence such that
∑n=1∞γn=∞;
limsupn→∞δn/γn≤0 or ∑n=1∞|δn|<∞.
Then limn→∞an=0.
3. Main Result
In this section, we will prove our main result. First, we give some assumptions on the operators and the parameters. Subsequently, we introduce our iterative algorithm for finding solutions of the variational inclusion (1.1). Finally, we will show that the proposed algorithm has strong convergence.
In the sequel, we will assume that
C is a nonempty closed convex subset of a real Hilbert space H;
A is a strongly positive bounded linear operator with coefficient 0<μ<1, R:H→2H is a maximal monotone mapping and B:C→C is an α-inverse strongly monotone mapping;
λ>0 is a constant satisfying λ<2α.
Now we introduce the following iteration algorithm.
Algorithm 3.1.
For given x0∈C arbitrarily, compute the sequence {xn} as follows:
xn+1=(1-βn)xn+βnPC[(I-αnA)JR,λ(I-λB)xn],n≥0,
where {αn} and {βn} are two real sequences in [0,1].
Now we study the strong convergence of the algorithm (3.1)
Theorem 3.2.
Suppose that I(B,R)≠∅. Assume the following conditions are satisfied:
limn→∞αn=0;
∑n=0∞αn=∞;
0<liminfn→∞βn≤limsupn→∞βn<1.
Then the sequence {xn} generated by (3.1) converges strongly to x̃∈I(B,R) which solves the following variational inequality:
〈Ax,y-x〉≥0,∀y∈I(B,R).
Proof.
Take x*∈I(B,R). It is clear that
JR,λ(x*-λBx*)=x*.
We divide our proofs into the following five steps:
the sequence {xn} is bounded.
∥xn+1-xn∥→0.
∥Bxn-Bx*∥→0.
limsupn→∞〈Ax̃,xn-x̃〉≥0 where x̃=PI(B,R)(I-A)(x̃).
xn→x̃.
Proof of (1.1).
Since B is α-inverse strongly monotone, we have
‖(I-λB)x-(I-λB)y‖2≤‖x-y‖2+λ(λ-2α)‖Bx-By‖2.
It is clear that if 0≤λ≤2α, then (I-λB) is nonexpansive. Set yn=JR,λ(xn-λBxn),n≥0. It follows that
‖yn-x*‖=‖JR,λ(xn-λBxn)-JR,λ(x*-λBx*)‖≤‖(xn-λBxn)-(x*-λBx*)‖≤‖xn-x*‖.
Since A is linear bounded self-adjoint operator on H, then
‖A‖=sup{|〈Au,u〉|:u∈H,‖u‖=1}.
Observe that
〈(I-αnA)u,u〉=1-αn〈Au,u〉≥1-αn‖A‖≥0,
that is to say I-αnA is positive. It follows that
‖(I-αnA)‖=sup{〈(I-αnA)u,u〉:u∈H,‖u‖=1}=sup{1-αn〈Au,u〉:u∈H,‖u‖=1}≤1-αnμ.
From (3.1), we deduce that
‖xn+1-x*‖=‖(1-βn)xn+βnPC[(I-αnA)yn]-x*‖≤(1-βn)‖xn-x*‖+βn‖[(I-αnA)(yn-x*)]-αnAx*‖≤(1-βn)‖xn-x*‖+(1-αnμ)βn‖yn-x*‖+αnβn‖Ax*‖≤(1-αnβnμ)‖xn-x*‖+αnβn‖Ax*‖≤max{‖x0-x*‖,‖Ax*‖μ}.
Therefore, {xn} is bounded.
Proof of (3.1).
Set zn=PC[(I-αnA)JR,λ(I-λB)xn] for all n≥0. Then, we have
‖zn-zn-1‖=‖PC[(I-αnA)yn]-PC[(I-αn-1A)yn-1]‖≤‖[(I-αnA)yn]-[(I-αn-1A)yn-1]‖=‖(I-αnA)(yn-yn-1)+(αn-1-αn)Ayn-1‖≤‖(I-αnA)(yn-yn-1)‖+‖(αn-1-αn)Ayn-1‖≤(1-αnμ)‖yn-yn-1‖+|αn-αn-1|‖Ayn-1‖.
Note that
‖yn-yn-1‖=‖JR,λ(xn-λBxn)-JR,λ(xn-1-λBxn-1)‖≤‖(xn-λBxn)-(xn-1-λBxn-1)‖≤‖xn-xn-1‖.
Substituting (3.11) into (3.10), we get
‖zn-zn-1‖≤(1-αnμ)‖xn-xn-1‖+|αn-αn-1|‖Ayn-1‖.
Therefore,
limsupn→∞(‖zn-zn-1‖-‖xn-xn-1‖)≤0.
This together with Lemma 2.2 imply that
limn→∞‖zn-xn‖=0.
Hence,
limn→∞‖xn+1-xn‖=limn→∞βn‖zn-xn‖=0.
Proof of (3.4).
From (3.4), we get
‖yn-x*‖2=‖JR,λ(xn-λBxn)-JR,λ(x*-λBx*)‖2≤‖(xn-λBxn)-(x*-λBx*)‖2≤‖xn-x*‖2+λ(λ-2α)‖Bxn-Bx*‖2.
By (3.1), we obtain
‖xn+1-x*‖2≤(1-βn)‖xn-x*‖2+βn‖PC[(I-αnA)yn]-x*‖2≤(1-βn)‖xn-x*‖2+βn‖(I-αnA)yn-x*‖2=(1-βn)‖xn-x*‖2+βn(‖yn-x*-αnAyn‖2)=(1-βn)‖xn-x*‖2+βn(‖yn-x*‖2-2αn〈yn-x*,Ayn〉+αn2‖Ayn‖2)≤(1-βn)‖xn-x*‖2+βn(‖yn-x*‖2+αn(2‖yn-x*‖‖Ayn‖+‖Ayn‖2))≤(1-βn)‖xn-x*‖2+βn(‖yn-x*‖2+αnM),
where M>0 is some constant satisfying supn{2∥yn-x*∥∥Ayn∥+∥Ayn∥2}≤M. From (3.16) and (3.17), we have
‖xn+1-x*‖2≤‖xn-x*‖2+λ(λ-2α)βn‖Bxn-Bx*‖2+αnM.
Thus,
λ(2α-λ)βn‖Bxn-Bx*‖2≤‖xn-x*‖2-‖xn+1-x*‖2+αnM≤(‖xn-x*‖+‖xn+1-x*‖)‖xn+1-xn‖+αnM,
which imply that
limn→∞‖Bxn-Bx*‖=0.
Proof of (3.10).
Since JR,λ is 1-inverse strongly monotone, we have
‖yn-x*‖2=‖JR,λ(xn-λBxn)-JR,λ(x*-λBx*)‖2≤〈xn-λBxn-(x*-λBx*),yn-x*〉=12(‖xn-λBxn-(x*-λBx*)‖2+‖yn-x*‖2-‖xn-λBxn-(x*-λBx*)-(yn-x*)‖2)≤12(‖xn-x*‖2+‖yn-x*‖2-‖xn-yn-λ(Bxn-Bx*)‖2)=12(‖xn-x*‖2+‖yn-x*‖2-‖xn-yn‖2+2λ〈Bxn-Bx*,xn-yn〉-λ2‖Bxn-Bx*‖2),
which implies that
‖yn-x*‖2≤‖xn-x*‖2-‖xn-yn‖2+2λ‖Bxn-Bx*‖‖xn-yn‖.
Substitute (3.22) into (3.17) to get
‖xn+1-x*‖2≤‖xn-x*‖2-βn‖xn-yn‖2+2λ‖Bxn-Bx*‖‖xn-yn‖+αnM.
Then we derive
βn‖xn-yn‖2≤(‖xn-x*‖+‖xn+1-x*‖)‖xn+1-xn‖+2λ‖Bxn-Bx*‖‖xn-yn‖+αnM.
So, we have
limn→∞‖xn-yn‖=0.
We note that PI(B,R)(I-A) is a contraction. As a matter of fact,
‖PI(B,R)(I-A)x-PI(B,R)(I-A)y‖≤‖(I-A)x-(I-A)y‖≤‖I-A‖‖x-y‖≤(1-μ)‖x-y‖,
for all x,y∈H. Hence PI(B,R)(I-A) has a unique fixed point, say x̃∈I(B,R). That is x̃=PI(B,R)(I-A)(x̃). This implies that 〈Ax̃,y-x̃〉≥0 for all y∈I(B,R). Next, we prove that
limsupn→∞〈Ax̃,xn-x̃〉≥0.
First, we note that there exists a subsequence {xni} of {xn} such that
limsupn→∞〈Ax̃,xn-x̃〉=limj→∞〈Ax̃,xnj-x̃〉.
Since {xnj} is bounded, there exists a subsequence {xnji} of {xnj} which converges weakly to w. Without loss of generality, we can assume that xnj⇀w.
Next, we show that w∈I(B,R). In fact, since B is α-inverse strongly monotone, B is Lipschitz-continuous monotone mapping. It follows from Lemma 2.1 that R+B is maximal monotone. Let (v,g)∈G(R+B), that is, g-Bv∈R(v). Again since yni=JR,λ(xni-λBxn-i), we have xni-λBxni∈(I+λR)(yni), that is, (1/λ)(xni-yni-λBxni)∈R(yni). By virtue of the maximal monotonicity of R+B, we have
〈v-yni,g-Bv-1λ(xni-yni-λBxni)〉≥0,
and so
〈v-yni,g〉≥〈v-yni,Bv+1λ(xni-yni-λBxni)〉=〈v-yni,Bv-Byni+Byni-Bxni+1λ(xni-yni)〉≥〈v-yni,Byni-Bxni〉+〈v-yni,1λ(xni-yni)〉.
It follows from ∥xn-yn∥→0, ∥Bxn-Byn∥→0 and yni⇀w that
limni→∞〈v-yni,g〉=〈v-w,g〉≥0.
It follows from the maximal monotonicity of B+R that θ∈(R+B)(w), that is, w∈I(B,R). Therefore, w∈I(B,R). It follows that
limsupn→∞〈Ax̃,xn-x̃〉=limj→∞〈Ax̃,xnj-x̃〉=〈Ax̃,w-x̃〉≥0.
Proof of (3.11).
First, we note that zn=PC[(I-αnA)yn], then for all x∈C, we have 〈zn-(I-αnA)yn,zn-x〉≤0. Thus,
‖zn-x̃‖2=〈zn-x̃,zn-x̃〉=〈zn-(I-αnA)yn+(I-αnA)yn-x̃,zn-x̃〉=〈zn-(I-αnA)yn,zn-x̃〉+〈(I-αnA)yn-x̃,zn-x̃〉≤〈(I-αnA)(yn-x̃)-αnAx̃,zn-x̃〉=〈(I-αnA)(yn-x̃),zn-x̃〉+αn〈-Ax̃,zn-x̃〉≤‖(I-αnA)(yn-x̃)‖‖zn-x̃‖+αn〈-Ax̃,zn-x̃〉≤(1-αnμ)2(‖xn-x̃‖2+‖zn-x̃‖2)+αn〈-Ax̃,zn-x̃〉,
that is,
‖zn-x̃‖2≤(1-αnμ)‖xn-x̃‖2+αn1+αnμ〈-Ax̃,zn-x̃〉.
So,
‖xn+1-x̃‖2≤(1-βn)‖xn-x*‖2+βn‖zn-x*‖2≤(1-βn)‖xn-x*‖2+βn(1-αnμ)‖xn-x̃‖2+αnβn1+αnμ〈-Ax̃,zn-x̃〉=(1-αnβnμ)‖xn-x*‖2+αnβn1+αnμ〈-Ax̃,zn-x̃〉=(1-δn)‖xn-x*‖2+δnσn,
where δn=αnβnμ and σn=(1/(1+αnμ)μ)〈-Ax̃,zn-x̃〉. It is easy to see that ∑n=1∞δn=∞ and limsupn→∞σn≤0. Hence, by Lemma 2.3, we conclude that the sequence {xn} converges strongly to x̃. This completes the proof.
4. Conclusion
The results proved in this paper may be extended for multivalued variational inclusions and related optimization problems.
Acknowledgment
This research was partially supported by Youth Foundation of Taizhou University (2011QN11).
RobinsonS. M.Generalized equations and their solutions. I: basic theory197910128141527064ZBL0404.90093RockafellarR. T.Monotone operators and the proximal point algorithm1976145877898041048310.1137/0314056ZBL0358.90053ZhangS. S.LeeJ. H. W.ChanC. K.Algorithms of common solutions to quasi variational inclusion and fixed point problems200829557158110.1007/s10483-008-0502-y2414681ZBL1196.47047PengJ. W.WangY.ShyuD. S.YaoJ. C.Common solutions of an iterative scheme for variational inclusions, equilibrium problems, and fixed point problems20082008157203712481571ZBL1161.65050NoorM. A.Equivalence of variational inclusions with resolvent equations2000417-896397010.1016/S0362-546X(98)00321-61764030ZBL0978.49010NoorM. A.RassiasT. M.Projection methods for monotone variational inequalities1999237240541210.1006/jmaa.1999.64221710303ZBL0933.65076NoorM. A.HuangZ.Some resolvent iterative methods for variational inclusions and nonexpansive mappings2007194126727510.1016/j.amc.2007.04.0372385850ZBL1193.49010NoorM. A.NoorK. I.Al-SaidE.Some resolvent methods for general variational
inclusions20112315361NoorM. A.NoorK. I.Sensitivity analysis for quasi-variational inclusions1999236229029910.1006/jmaa.1999.64241704584ZBL0949.49007NoorM. A.Three-step iterative algorithms for multivalued quasi variational inclusions2001255258960410.1006/jmaa.2000.72981815802ZBL0986.49006NoorM. A.Generalized set-valued variational inclusions and resolvent equations1998228120622010.1006/jmaa.1998.61271659909ZBL1031.49016YaoY.ChoY. J.LiouY. C.Algorithms of common solutions for variational inclusions, mixed equilibrium problems and fixed point problems2011212224225010.1016/j.ejor.2011.01.0422784202YaoY.ShahzadN.New methods with perturbations for non-expansive mappings in
Hilbert spaces20112011, article ID 7910.1186/1687-1812-2011-79YaoY.ShahzadN.Strong convergence of a proximal point algorithm with general
errorsOptimization Letters. In press10.1007/s11590-011-0286-2YaoY.LiouY. C.ChenC. P.Algorithms construction for nonexpansive mappings and inverse-strongly monotone mappings20111519791998YaoY.ChenR.LiouY. C.A unified implicit algorithm for solving the triple-hierarchical constrained optimization problemMathematical & Computer Modelling. In pressAdlyS.Perturbed algorithms and sensitivity analysis for a general class of variational inclusions1996201260963010.1006/jmaa.1996.02771396922ZBL0856.65077ChangS. S.Set-valued variational inclusions in Banach spaces2000248243845410.1006/jmaa.2000.69191776022ZBL1031.49018ChangS. S.Existence and approximation of solutions of set-valued variational inclusions in Banach spaces200147583594DingX. P.Perturbed Ishikawa type iterative algorithm for generalized quasivariational inclusions20031412-3359373197291610.1016/S0096-3003(02)00261-8ZBL1030.65071HuangN. J.Mann and Ishikawa type perturbed iterative algorithms for generalized nonlinear implicit quasi-variational inclusions199835101710.1016/S0898-1221(98)00066-21617871ZBL0999.47057LinL. J.Variational inclusions problems with applications to Ekeland's variational principle, fixed point and optimization problems200739450952710.1007/s10898-007-9153-12350374ZBL1190.90212NoorM. A.Some developments in general variational inequalities20041521199277205006310.1016/S0096-3003(03)00558-7ZBL1134.49304NoorM. A.Extended general variational inequalities200922218218610.1016/j.aml.2008.03.0072482273ZBL1163.49303NoorM. A.General variational inequalities19881211912110.1016/0893-9659(88)90054-7953368ZBL0655.49005YaoY.LiouY. C.KangS. M.Two-step projection methods for a system of variational inequality problems in Banach spacesJournal of Global Optimization. In pressYaoY.NoorM. A.LiouY. C.Strong convergence of a modied extra-gradient method to the minimum-norm solution of variational inequalities20122012981743610.1155/2012/817436YaoY.LiouY. C.LiC. L.LinH. T.Extended extra-gradient methods for generalized variational inequalities201220121423708310.1155/2012/237083LemaireB.Which fixed point does the iteration method select?1997452Berlin, GermanySpringerLecture Notes in Economics and Mathematical Systems1467027ZBL0882.65042BrezisH.1973Amsterdam, The NetherlandsNorth-Holland Publishing0348562XuH. K.Iterative algorithms for nonlinear operators200266124025610.1112/S00246107020033321911872ZBL1013.47032