AAAAbstract and Applied Analysis1687-04091085-3375Hindawi Publishing Corporation85436010.1155/2011/854360854360Review ArticleThe General Hybrid Approximation Methods for Nonexpansive Mappings in Banach SpacesWangkeereeRabianKocinacLjubisaDepartment of MathematicsFaculty of ScienceNaresuan UniversityPhitsanulok 65000Thailandnu.ac.th201122022011201123112010270120112011Copyright © 2011 Rabian Wangkeeree.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 introduce two general hybrid iterative approximation methods (one implicit and one explicit) for finding a fixed point of a nonexpansive mapping which solving the variational inequality generated by two strongly positive bounded linear operators. Strong convergence theorems of the proposed iterative methods are obtained in a reflexive Banach space which admits a weakly continuous duality mapping. The results presented in this paper improve and extend the corresponding results announced by Marino and Xu (2006), Wangkeeree et al. (in press), and Ceng et al. (2009).

1. Introduction

Let C be a nonempty subset of a normed linear space E. Recall that a mapping T:CC is called nonexpansive ifTx-Tyx-y,x,yE. We use F(T) to denote the set of fixed points of T; that is, F(T)={xE:Tx=x}. A self-mapping f:EE is a contraction on E if there exists a constant α(0,1) and x,yE such thatf(x)-f(y)αx-y.

One classical way to study nonexpansive mappings is to use contractions to approximate a nonexpansive mapping . More precisely, take t(0,1) and define a contraction Tt:EE byTtx=tu+(1-t)Tx,xE, where uE is a fixed point. Banach's contraction mapping principle guarantees that Tt has a unique fixed point xt in E. It is unclear, in general, what is the behavior of xt as t0, even if T has a fixed point. However, in the case of T having a fixed point, Browder  proved that if E is a Hilbert space, then xt converges strongly to a fixed point of T. Reich  extended Browder's result to the setting of Banach spaces and proved that if E is a uniformly smooth Banach space, then {xt} converges strongly to a fixed point of T and the limit defines the (unique) sunny nonexpansive retraction from E onto F(T). Xu  proved Reich's results hold in reflexive Banach spaces which have a weakly continuous duality mapping.

The iterative methods for nonexpansive mappings have recently been applied to solve convex minimization problems; see, for example,  and the references therein. Let H be a real Hilbert space, whose inner product and norm are denoted by ·,· and ·, respectively. Let A be a strongly positive bounded linear operator on H: that is, there is a constant γ¯>0 with propertyAx,xγ¯x2,xH. A typical problem is to minimize a quadratic function over the set of the fixed points of a nonexpansive mapping on a real Hilbert space HminxF(T)12Ax,x-x,b, where T is a nonexpansive mapping on H and b is a given point in H. In 2003, Xu  proved that the sequence {xn} defined by the iterative method below, with the initial guess x0H chosen arbitrarilyxn+1=(I-λnA)Txn+λnu,n0 converges strongly to the unique solution of the minimization problem (1.5) provided the sequence {λn} satisfies certain conditions. Using the viscosity approximation method, Moudafi  introduced the following iterative iterative process for nonexpansive mappings (see [9, 10] for further developments in both Hilbert and Banach spaces). Let f be a contraction on H. Starting with an arbitrary initial x0H, define a sequence {xn} recursively byxn+1=(1-λn)Txn+λnf(xn),n0, where {λn} is a sequence in (0,1). It is proved [8, 10] that under certain appropriate conditions imposed on {λn}, the sequence {xn} generated by (1.7) strongly converges to the unique solution x* in C of the variational inequality(I-f)x*,x-x*0,xH. Recently, Marino and Xu  mixed the iterative method (1.6) and the viscosity approximation method (1.7) and considered the following general iterative method:xn+1=(I-λnA)Txn+λnγf(xn),n0, where A is a strongly positive bounded linear operator on H. They proved that if the sequence {λn} of parameters satisfies the following appropriate conditions: limnλn    =0, n=1λn=, and either n=1|λn+1-λn|< or limn(λn/λn+1)=1, then the sequence {xn} generated by (1.9) converges strongly to the unique solution x* in H of the variational inequality(A-γf)x*,x-x*0,xH, which is the optimality condition for the minimization problem: minxF(T)(1/2)Ax,x-h(x), where h is a potential function for γf(i.e.,h(x)=γf(x) for xH).

Very recently, Wangkeeree et al.  extended Marino and Xu's result to the setting of Banach spaces and obtained the strong convergence theorems in a reflexive Banach space which admits a weakly continuous duality mapping. Let E be a reflexive Banach space which admits a weakly continuous duality mapping Jφ with gauge φ such that φ is invariant on [0,1]. Let T:EE be a nonexpansive mapping with F(T), f a contraction with coefficient 0<α<1 and A a strongly positive bounded linear operator with coefficient γ¯>0 and 0<γ<γ¯φ(1)/α. Define the net {xt} byxt=tγf(xt)+(I-tA)Txt. It is proved in  that {xt} converges strongly as t0 to a fixed point x̃ of T which solves the variational inequality(A-γf)x̃,Jφ(x̃-z)0,zF(T).

On the other hand, Ceng et al.  introduced the iterative approximation method for solving the variational inequality generated by two strongly positive bounded linear operators on a real Hilbert space H. Let f:HH be a contraction with coefficient 0<α<1, and let A,B:HH be two strongly positive bounded linear operators with coefficient γ¯(0,1) and β>0, respectively. Assume that 0<γα<β, {λn} is a sequence in (0,1), {μn} is a sequence in (0,min{1,B-1}). Starting with an arbitrary initial x0H, define a sequence {xn} recursively byxn+1=(1-λnA)Txn+λn+1[Txn-μn+1(BTxn-γf(xn))],n0. It is proved in [13, Theorem 3.1] that if the sequences {λn} and {μn} satisfy the following conditions:

limnλn=0,

n=1λn=,

n=1|λn+1-λn|< or limn(λn/λn+1)=1,

(1-γ¯)/(β-γα)<limnμn=μ<(2-γ¯)/(β-γα),

then the sequence {xn} generated by (1.13) converges strongly to the unique solution x̃ in H of the variational inequality(A-I+μ(B-γf))x̃,x̃-z0,zF(T). Observe that if B=I and μn=1 for all n1, then algorithm (1.13) reduces to (1.9). Moreover, the variational inequality (1.14) reduces to (1.10). Furthermore, the applications of these results to constrained generalized pseudoinverse are studied.

In this paper, motivated by Marino and Xu , Wangkeeree et al. , and Ceng et al. , we introduce two general iterative approximation methods (one implicit and one explicit) for finding a fixed point of a nonexpansive mapping which solving the variational inequality generated by two strongly positive bounded linear operators. Strong convergence theorems of the proposed iterative methods are obtained in a reflexive Banach space which admits a weakly continuous duality mapping. The results presented in this paper improve and extend the corresponding results announced by Marino and Xu , Wangkeeree et al. , and Ceng et al. , and many others.

2. Preliminaries

Throughout this paper, let E be a real Banach space and E* its dual space. We write xnx (resp. xn*x) to indicate that the sequence {xn} weakly (resp. weak*) converges to x; as usual, xnx will symbolize strong convergence. Let UE={xE:x=1}. A Banach space E is said to uniformly convex if, for any ɛ(0,2], there exists δ>0 such that, for any x,yUE, x-yɛ implies (x+y)/21-δ. It is known that a uniformly convex Banach space is reflexive and strictly convex (see also ). A Banach space E is said to be smooth if the limit limt0(x+ty-x)/t exists for all x,yUE. It is also said to be uniformly smooth if the limit is attained uniformly for x,yUE.

By a gauge function φ, we mean a continuous strictly increasing function φ:[0,)[0,) such that φ(0)=0 and φ(t) as t. Let E* be the dual space of E. The duality mapping Jφ:E2E* associated to a gauge function φ is defined by Jφ(x)={f*E*:x,f*=xφ(x),f*=φ(x)},xE.

In particular, the duality mapping with the gauge function φ(t)=t, denoted by J, is referred to as the normalized duality mapping. Clearly, there holds the relation Jφ(x)=(φ(x)/x)J(x) for all x0 (see ). Browder  initiated the study of certain classes of nonlinear operators by means of the duality mapping Jφ. Following Browder , we say that a Banach space E has a weakly continuous duality mapping if there exists a gauge φ for which the duality mapping Jφ(x) is single valued and continuous from the weak topology to the weak topology; that is, for any {xn} with xnx, the sequence {Jφ(xn)} converges weakly to Jφ(x). It is known that lp has a weakly continuous duality mapping with a gauge function φ(t)=tp-1 for all 1<p<. Set Φ(t)=0tφ(τ)dτ,t0,

then Jφ(x)=Φ(x),xE,

where denotes the subdifferential in the sense of convex analysis.

Now, we collect some useful lemmas for proving the convergence result of this paper.

The first part of the next lemma is an immediate consequence of the subdifferential inequality, and the proof of the second part can be found in .

Lemma 2.1 (see [<xref ref-type="bibr" rid="B5">16</xref>]).

Assume that a Banach space E has a weakly continuous duality mapping Jφ with gauge φ.

For all x,yE, the following inequality holds:

Φ(x+y)Φ(x)+y,Jφ(x+y). In particular, for all x,yE, x+y2x2+2y,J(x+y).

Assume that a sequence {xn} in E converges weakly to a point xE. Then, the following identity holds:

limsupnΦ(xn-y)=limsupnΦ(xn-x)+Φ(y-x),x,yE.

Lemma 2.2 (see [<xref ref-type="bibr" rid="B14">7</xref>]).

Assume that {an} is a sequence of nonnegative real numbers such that an+1(1-αn)an+bn, where {αn} is a sequence in (0,1) and {bn} is a sequence such that

n=1αn=,

limsupnbn/αn0 or n=1|bn|<.

Then, limnan=0.

In a Banach space E having a weakly continuous duality mapping Jφ with a gauge function φ, an operator A is said to be strongly positive  if there exists a constant γ¯>0 with the propertyAx,Jφ(x)γ¯xφ(x),αI-βA=supx1|(αI-βA)x,Jφ(x)|,α[0,1],β[-1,1], where I is the identity mapping. If E:=H is a real Hilbert space, then (2.8) reduces to (1.4). The next valuable lemma can be found in .

Lemma 2.3 (see [<xref ref-type="bibr" rid="B12">12</xref>, Lemma 3.1]).

Assume that a Banach space E has a weakly continuous duality mapping Jφ with gauge φ. Let A be a strongly positive bounded linear operator on E with coefficient γ¯>0 and 0<ρφ(1)A-1. Then, I-ρAφ(1)(1-ργ¯).

3. Main Results

Now, we are a position to state and prove our main results.

Lemma 3.1.

Let E be a Banach space which admits a weakly continuous duality mapping Jφ with gauge φ such that φ is invariant on [0,1]; that is, T([0,1])[0,1]. Let T:EE be a nonexpansive mapping and f:EE a contraction with coefficient α(0,1). Let A and B be two strongly positive bounded linear operators with coefficients γ¯>0 and β>0, respectively. Let γ and μ be two constants satisfying the condition (C*)(C*):0<γ<βφ(1)α,φ(1)-φ(1)γ¯φ(1)β-γα<μmin{1,φ(1)B-1,1+φ(1)-φ(1)γ¯φ(1)β-γα}. Then, for any λ(0,min{1,φ(1)A-1}), the mapping Sλ:EE defined by Sλ(x)=(I-λA)Tx+λ[Tx-μ(BTx-γf(x))],xE is a contraction with coefficient 1-λτ, where τ:=φ(1)γ¯-φ(1)+μ(φ(1)β-γα).

Proof.

Observe that μ1+φ(1)-φ(1)γ¯φ(1)β-γαμ(φ(1)β-γα)1+φ(1)-φ(1)γ¯φ(1)γ¯-φ(1)+μ(φ(1)β-γα)1,φ(1)-φ(1)γ¯φ(1)β-γα<μφ(1)-φ(1)γ¯<μ(φ(1)β-γα)0<φ(1)γ¯-φ(1)+μ(φ(1)β-γα). This shows that τ=φ(1)γ¯-φ(1)+μ(φ(1)β-γα)(0,1]. Using Lemma 2.3, we obtain Sλ(x)-Sλ(y)=(I-λA)Tx+λ[Tx-μ(BTx-γf(x))]-(I-λA)Ty-λ[Ty-μ(BTy-γf(y))](I-λA)Tx-(I-λA)Ty+λTx-μ(BTx-γf(x))-[Ty-μ(BTy-γf(y))]I-λATx-Ty+λ[(I-μB)Tx-(I-μB)Ty+γμf(x)-f(y)]I-λATx-Ty+λ[I-μBTx-Ty+γμf(x)-f(y)]φ(1)(1-λγ¯)x-y+λ[φ(1)(1-μβ)x-y+γμαx-y]=[φ(1)(1-λγ¯)+λ[φ(1)(1-μβ)+γμα]]x-y=[φ(1)(1-λγ¯)+λ[φ(1)-μ(φ(1)β-γα)]]x-y=[φ(1)-λ[φ(1)γ¯-φ(1)+μ(φ(1)β-γα)]]x-y=(φ(1)-λτ)x-y(1-λτ)x-y. Hence, Sλ is a contraction with coefficient 1-λτ.

Applying the Banach contraction principle to Lemma 3.1, there exists a unique fixed point xλ of Sλ in E; that is,xλ=(I-λA)Txλ+λ[Txλ-μ(BTxλ-γf(xλ))],λ(0,1).

Remark 3.2.

For each 1<p<, lp space has a weakly continuous duality mapping with a gauge function φ(t)=tp-1 which is invariant on [0,1].

Theorem 3.3.

Let E be a reflexive Banach space which admits a weakly continuous duality mapping Jφ with gauge φ such that φ is invariant on [0,1]. Let T:EE be a nonexpansive mapping with F(T), f:EE a contraction with coefficient α(0,1), and A, B two strongly positive bounded linear operators with coefficients γ¯>0 and β>0, respectively. Let γ and μ be two constants satisfying the condition (C*). Then, the net {xλ} defined by (3.5) converges strongly as λ0 to a fixed point x̃ of T which solves the variational inequality (A-I+μ(B-γf))x̃,Jφ(x̃-z)0,zF(T).

Proof.

We first show that the uniqueness of a solution of the variational inequality (3.6). Suppose that both x̃F(T) and x*F(T) are solutions to (3.6), then (A-I+μ(B-γf))x̃,Jφ(x̃-x*)0,(A-I+μ(B-γf))x*,Jφ(x*-x̃)0. Adding (3.7), we obtain (A-I+μ(B-γf))x̃-(A-I+μ(B-γf))x*,Jφ(x̃-x*)0. On the other hand, we observe that φ(1)-φ(1)γ¯φ(1)β-γα<μφ(1)-φ(1)γ¯<μ(φ(1)β-γα)1-γ¯<μ(β-γαφ(1))0<γ¯-1+μ(β-γαφ(1)). It then follows that for any x,yE, (A-I+μ(B-γf))x-(A-I+μ(B-γf))y,Jφ(x-y)=A(x-y)-(x-y)+μ[(B-γf)x-(B-γf)y],Jφ(x-y)=A(x-y),Jφ(x-y)-x-y,Jφ(x-y)+  μ(B-γf)x-(B-γf)y,Jφ(x-y)γ¯x-yφ(x-y)-x-yφ(x-y)+μB(x-y),Jφ(x-y)-μγf(x)-f(y),Jφ(x-y)γ¯x-yφ(x-y)-x-yφ(x-y)+μβx-yφ(x-y)-μγf(x)-f(y)Jφ(x-y)γ¯Φ(x-y)-Φ(x-y)+μβΦ(x-y)-μγαΦ(x-y)=(γ¯-1+μβ-μγα)γ¯Φ(x-y)=(γ¯-1+μ(β-γα))γ¯Φ(x-y)(γ¯-1+μ(β-γαφ(1)))γ¯Φ(x-y)0. Applying (3.10) to (3.8), we obtain that x̃=x* and the uniqueness is proved. Below, we use x̃ to denote the unique solution of (3.6). Next, we will prove that {xλ} is bounded. Take a pF(T), and denote the mapping Sλ by Sλ=(I-λA)T+λ[T-μ(BT-γf)],λ(0,1). From Lemma 3.1, we have xλ-pSλxλ-Sλp+Sλp-p(1-λτ)xλ-p+(I-λA)Tp+λ[Tp-μ(BTp-γfp)]-p=(1-λτ)xλ-p+λ-Ap+p-μ(Bp-γfp)(1-λτ)xλ-p+λ[I-Ap+μBp-γfp], where τ:=φ(1)γ¯-φ(1)+μ(φ(1)β-γα)(0,1]. It follows that xλ-p1τ[I-Ap+μBp-γfp]. Hence, {xλ} is bounded, so are {f(xλ)}, {AT(xλ)} and {BT(xλ)}. The definition of {xλ} implies that xλ-Txλ=λTxλ-μ(BTxλ-γf(xλ))-ATxλ0,as  λ0. If follows from reflexivity of E and the boundedness of sequence {xλ} that there exists {xλn} which is a subsequence of {xλ} converging weakly to wE as n. Since Jφ is weakly sequentially continuous, we have by Lemma 2.1 that limsupnΦ(xλn-x)=limsupnΦ(xλn-w)+Φ(x-w),xE. Let H(x)=limsupnΦ(xλn-x),xE. It follows that H(x)=H(w)+Φ(x-w),xE. Since xλn-Txλn=λnTxλn-μ(BTxλn-γf(xλn))-ATxλn0,as  n. We obtain H(Tw)=limsupnΦ(xλn-Tw)=limsupnΦ(Txλn-Tw)limsupnΦ(xλn-w)=H(w). On the other hand, however, H(Tw)=H(w)+Φ(T(w)-w). It follows from (3.19) and (3.20) that Φ(T(w)-w)=H(Tw)-H(w)0, which gives us, Tw=w. Next, we show that xλnw as n. In fact, since Φ(t)=0tφ(τ)dτ,    t0 and φ:[0,)[0,) is a gauge function, then for 1k0, φ(kx)φ(x) and Φ(kt)=0ktφ(τ)dτ=k0tφ(kx)dxk0tφ(x)dx=kΦ(t). Following Lemma 2.1, we have Φ(xλn-w)=Φ((I-λnA)Txλn+λn[Txλn-μ(BTxλn-γf(xλn))]-(I-λnA)w-λnAw)Φ((I-λnA)Txλn-(I-λnA)w)+  λnTxλn-μ(BTxλn-γf(xλn))-Aw,Jφ(xλn-w)Φ(φ(1)(1-λnγ¯)xλn-w)+  λn(I-μB)Txλn+μγf(xλn)-Aw,Jφ(xλn-w)φ(1)(1-λnγ¯)Φ(xλn-w)+  λn(I-μB)Txλn-(I-μB)w+μγf(xλn)-μγf(w),Jφ(xλn-w)+  λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w)φ(1)(1-λnγ¯)Φ(xλn-w)+λn(I-μB)Txλn-(I-μB)w,Jφ(xλn-w)+  λnμγf(xλn)-f(w),Jφ(xλn-w)+λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w)φ(1)(1-λnγ¯)Φ(xλn-w)+λn(I-μB)Txλn-(I-μB)wJφ(xλn-w)+  λnμγf(xλn)-f(w)Jφ(xλn-w)+λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w)φ(1)(1-λnγ¯)Φ(xλn-w)+λnφ(1)(1-μβ)xλn-wJφ(xλn-w)+  λnμγαxλn-wJφ(xλn-w)+λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w)=[φ(1)(1-λnγ¯)+λn(φ(1)(1-μβ)+μγα)]Φ(xλn-w)+  λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w)=[φ(1)-λn(φ(1)γ¯-φ(1)+μ(φ(1)β-γα))]Φ(xλn-w)+  λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w)[1-λn(φ(1)γ¯-φ(1)+μ(φ(1)β-γα))]Φ(xλn-w)+  λn(I-μB)w+μγf(w)-Aw,Jφ(xλn-w). Thus, Φ(xλn-w)1φ(1)γ¯-φ(1)+μ(φ(1)β-γα)(I-μB)w+μγf(w)-Aw,Jφ(xλn-w). Now, observing that xλnw implies Jφ(xλn-w)0, we conclude from the last inequality that Φ(xλn-w)0,as  n. Hence, xλnw as n. Next, we prove that w solves the variational inequality (3.6). For any zF(T), we observe that (I-T)xλ-(I-T)z,Jφ(xλ-z)=xt-z,Jφ(xλ-z)+Txt-Tz,Jφ(xλ-z)=Φ(xλ-z)-Tz-Txt,Jφ(xλ-z)Φ(xλ-z)-Tz-TxtJφ(xλ-z)Φ(xλ-z)-z-xtJφ(xλ-z)=Φ(xλ-z)-Φ(xλ-z)=0. Since xλ=(I-λnA)Txλn+λn[Txλn-μ(BTxλn-γf(xλn))], we can derive that λn[Axλn-(I-μB)xλn]=(I-λnA)Txλn-(I-λnA)xλn+λn(I-μB)Txλn-λn(I-μB)xλn+λnγf(xλn). That is [A-I+μ(B-γf)]xλn=-1λn[(I-λnA)(I-T)xλn+λn(I-μB)(I-T)xλn]. Using (3.26), for each pF(T), we have [A-I+μ(B-γf)]xλn,Jφ(xλn-p)=-1λn[(I-λnA)(I-T)xλn,Jφ(xλn-p)+λn(I-μB)(I-T)xλn,Jφ(xλn-p)]=-1λn(I-T)xλn-(I-T)p,Jφ(xλn-p)+A(I-T)xλn,Jφ(xλn-p)-(I-T)xλn-(I-T)p,Jφ(xλn-p)+μB(I-T)xλn,Jφ(xλn-p)A(I-T)xλn,Jφ(xλn-p)+B(I-T)xλn,Jφ(xλn-p)Axλn-TxλnJφ(xλn-p)+μBxλn-TxλnJφ(xλn-p)xλn-TxλnM, where M is a constant satisfying Msupn1{AJφ(xλn-p),μBJφ(xλn-p)}. Noticing that xλn-Txλnw-T(w)=w-w=0. It follows from (3.30) that (A-I+μ(B-γf))w,Jφ(w-p)0. So, wF(T) is a solution of the variational inequality (3.6), and hence, w=x̃ by the uniqueness. In a summary, we have shown that each cluster point of {xλ}(at λ0) equals x̃. Therefore, xλx̃ as λ0. This completes the proof.

According to the definition of strongly positive operator A in a Banach space E having a weakly continuous duality mapping Jφ with a gauge function φ, an operator A is said to be strongly positive  if there exists a constant γ¯>0 with the property Ax,Jφ(x)γ¯xφ(x),αI-βA=supx1|(αI-βA)x,Jφ(x)|,α[0,1],β[-1,1],

where I is the identity mapping. We may assume, without loss of generality, that γ¯<1. Therefore, if 0<γ<γ¯φ(1)/α, then we have the Corollary 3.4 immediately. Indeed, putting B=I and β=1, we haveφ(1)-φ(1)γ¯φ(1)β-γα=φ(1)-φ(1)γ¯φ(1)-γα<1<1+φ(1)-φ(1)γ¯φ(1)-γα=1+φ(1)-φ(1)γ¯φ(1)β-γα.

Taking μ1 in Theorem 3.3, we obtain the following result.

Corollary 3.4 (see [<xref ref-type="bibr" rid="B12">12</xref>, Lemma 3.3]).

Let E be a reflexive Banach space which admits a weakly continuous duality mapping Jφ with gauge φ such that φ is invariant on [0,1]. Let T:EE be a nonexpansive mapping with F(T), f:EE a contraction with coefficient α(0,1), and A a strongly positive bounded linear operator with coefficient γ¯>0 and 0<γ<γ¯φ(1)/α. Then, the net {xλ} defined by xλ=(I-λA)Txλ+λγf(xλ), converges strongly as λ0 to a fixed point x̃ of T which solves the variational inequality: (A-γf)x̃,Jφ(x̃-z)0,zF(T).

Corollary 3.5 (see [<xref ref-type="bibr" rid="B6">11</xref>, Theorem 3.6]).

Let H be a real Hilbert space. Let T:HH be a nonexpansive mapping with F(T), f:HH a contraction with coefficient α(0,1), and A a strongly positive bounded linear operator with coefficient γ¯>0 and 0<γ<γ¯/α. Then, the net {xλ} defined by xλ=(I-λA)Txλ+λγf(xλ), converges strongly as λ0 to a fixed point x̃ of T which solves the variational inequality (A-γf)x̃,x̃-z0,zF(T).

Theorem 3.6.

Let E be a reflexive Banach space which admits a weakly continuous duality mapping Jφ with gauge φ such that φ is invariant on [0,1]. Let T:EE be a nonexpansive mapping with F(T), f:EE a contraction with coefficient α(0,1), and A and B two strongly positive bounded linear operators with coefficients γ¯>0 and β>0, respectively. Let x0E be arbitrary and the sequence {xn} generated by the following iterative scheme: xn+1=(I-λnA)Txn+λn[Txn-μ(BTxn-γf(xn))],n0, where γ and μ are two constants satisfying the condition (C*) and {λn} is a real sequence in (0,1) satisfying the following conditions:

limnλn=0 and n=1λn=,

n=1|λn+1-λn|< or limn(λn/λn+1)=1.

Then, the sequence {xn} defined by (3.39) converges strongly to a fixed point x̃ of T that is obtained by Theorem 3.3.

Proof.

We first prove that {xn} is bounded. Take a pF(T), and denote Sλn=(I-λnA)T+λn[T-μ(BT-γf)]. Using Lemma 3.1, we have xn+1-pSλnxn-Sλnp+Sλnp-p(1-λnτ)xn-p+(I-λnA)Tp+λn[Tp-μ(BTp-γfp)]-p=(1-λnτ)xn-p+λn-Ap+p-μ(Bp-γfp)(1-λnτ)xn-p+λn[I-Ap+μBp-γfp]=(1-λnτ)xn-p+λnτ[I-Ap+μBp-γfp]τmax{xn-p,τ[I-Ap+μBp-γfp]τ}, where τ=φ(1)γ¯-φ(1)+μ(φ(1)β-γα)(0,1]. By induction, it is easy to see that xn-pmax{x0-p,τ[I-Ap+μBp-γfp]τ},n0. Thus, {xn} is bounded, and hence so are {yn}, {ATxn}, {BTxn}, and {f(xn)}. Now, we show that limnxn+1-xn=0. From the definition of {xn}, it is easily seen that Sλn+1xn-Sλnxn=(I-λn+1A)Txn+λn+1[Txn-μ(BTxn-γf(xn))]-(I-λnA)Txn-λn[Txn-μ(BTxn-γf(xn))]=(λn-λn+1)ATxn+(λn+1-λn)Txn+μ(λn-λn+1)(BTxn-γf(xn))=(λn+1-λn)(I-A)Txn+μ(λn-λn+1)(BTxn-γf(xn)). It follows that xn+2-xn+1=Sλn+1xn+1-SλnxnSλn+1xn+1-Sλn+1xn+Sλn+1xn-Sλnxn(1-λn+1τ)xn+1-xn+|λn+1-λn|(I-A)Txn+μ|λn-λn+1|BTxn-γf(xn)(1-λn+1τ)xn+1-xn+(1+μ)|λn+1-λn|M=(1-λn+1τ)xn+1-xn+(1+μ)λn+1τ|λn+1-λn|λn+1τM, where M is a constant satisfying Msup{(I-A)Txn,BTxn-γf(xn)}. From condition (C2), we deduce that either n=1|λn+1-λn|M< or limn((λn+1-λn)/λn+1τ)M=0. Therefore, it follows from Lemma 2.2 that limnxn+1-xn=0. It then follows that xn-Txnxn-xn+1+xn+1-Txn=xn-xn+1+λnTxn-μ(BTxn-γf(xn))-ATxn0. Next, we prove that limsupn-(A-I+μ(B-γf))x̃,Jφ(xn-x̃)0. Let {xnk} be a subsequence of {xn} such that limk-(A-I+μ(B-γf))x̃,Jφ(xnk-x̃)=limsupn-(A-I+μ(B-γf))x̃,Jφ(xn-x̃). If follows from reflexivity of E and the boundedness of a sequence {xnk} that there exists {xnki} which is a subsequence of {xnk} converging weakly to wE as i. Since Jφ is weakly continuous, we have by Lemma 2.1 that limsupnΦ(xnki-x)=limsupnΦ(xnki-w)+Φ(x-w),xE. Let H(x)=limsupnΦ(xnki-x),xE. It follows that H(x)=H(w)+Φ(x-w),xE. From (3.46), we obtain H(Tw)=limsupiΦ(xnki-Tw)=limsupiΦ(Txnki-Tw)limsupiΦ(xnki-w)=H(w). On the other hand, however, H(Tw)=H(w)+Φ(T(w)-w). It follows from (3.52) and (3.53) that Φ(T(w)-w)=H(Tw)-H(w)0. This implies that Tw=w. Since the duality map Jφ is single valued and weakly continuous, we get that limsupn-(A-I+μ(B-γf))x̃,Jφ(xn-x̃)=limk-(A-I+μ(B-γf))x̃,Jφ(xnk-x̃)=limi-(A-I+μ(B-γf))x̃,Jφ(xnki-x̃)=-(A-I+μ(B-γf))x̃,Jφ(w-x̃)=(A-I+μ(B-γf))x̃,Jφ(x̃-w)0 as required. Finally, we show that xnx̃ as nΦ(xn+1-x̃)=Φ((I-λnA)Txn+λn[Txn-μ(BTxn-γf(xn))]-(I-λnA)x̃-λnAx̃)Φ((I-λnA)Txn-(I-λnA)x̃)+λnTxn-μ(BTxn-γf(xn))-Ax̃,Jφ(xn+1-x̃)φ(1)(1-λnγ¯)Φ(xn-x̃)  +λn(I-μB)Txn+γμf(xn)  -Ax̃,Jφ(xn+1-x̃)=φ(1)(1-λnγ¯)Φ(xn-x̃)+λn[(I-μB)Txn+γμf(xn)-(I-μB)Txn+1-γμf(xn+1),Jφ(xn+1-x̃)+(I-μB)Txn+1+γμf(xn+1)-(I-μB)x̃-γμf(x̃),Jφ(xn+1-x̃)+(I-μB)x̃+γμf(x̃)-A(x̃),Jφ(xn+1-x̃))]=φ(1)(1-λnγ¯)Φ(xn-x̃)+λn[(I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)(I-μB)(Txn-Txn+1),Jφ(xn+1-x̃)+γμf(xn)-f(xn+1),Jφ(xn+1-x̃)+(I-μB)(Txn+1-x̃),Jφ(xn+1-x̃)+γμf(xn+1)-f(x̃),Jφ(xn+1-x̃)+I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)]=φ(1)(1-λnγ¯)Φ(xn-x̃)+λn[(I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)φ(1)(1-μβ)xn-xn+1Jφ(xn+1-x̃)+γμαxn-xn+1Jφ(xn+1-x̃)+φ(1)(1-μβ)xn+1-x̃Jφ(xn+1-x̃)+γμαxn+1-x̃Jφ(xn+1-x̃)+I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)]φ(1)(1-λnγ¯)Φ(xn-x̃)+λn[(I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)φ(1)(1-μβ)xn-xn+1M+γμαxn-xn+1M+I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)]+λn[φ(1)-μ(φ(1)β-γα)]Φ(xn+1-x̃), where M is a constant satisfying Msupn0Jφ(xn+1-x̃). It then follows that Φ(xn+1-x̃)φ(1)(1-λnγ¯)1-λn[φ(1)-μ(φ(1)β-γα)]Φ(xn-x̃)+λn[φ(1)(1-μβ)1-λn[φ(1)-μ(φ(1)β-γα)]xn-xn+1M+γμα1-λn[φ(1)-μ(φ(1)β-γα)]xn-xn+1M+11-λn[φ(1)-μ(φ(1)β-γα)]I-A-μ(B-γf)x̃,Jφ(xn+1-x̃)]=(1-λn[φ(1)γ¯-(φ(1)-μ(φ(1)β-γα))]1-λn[φ(1)-μ(φ(1)β-γα)])Φ(xn-x̃)+λn[φ(1)(1-μβ)1-λn[φ(1)-μ(φ(1)β-γα)]xn-xn+1M+γμα1-λn[φ(1)-μ(φ(1)β-γα)]xn-xn+1M+11-λn[φ(1)-μ(φ(1)β-γα)]-A-I+μ(B-γf)x̃,Jφ(xn+1-x̃)]. Put γn=λn[φ(1)γ¯-(φ(1)-μ(φ(1)β-γα))]1-λn[φ(1)-μ(φ(1)β-γα)],δn=1-λn[φ(1)-μ(φ(1)β-γα)][φ(1)γ¯-(φ(1)-μ(φ(1)β-γα))]×[φ(1)(1-μβ)1-λn[φ(1)-μ(φ(1)β-γα)]xn-xn+1M+γμα1-λn[φ(1)-μ(φ(1)β-γα)]xn-xn+1M+11-λn[φ(1)-μ(φ(1)β-γα)]-A-I+μ(B-γf)x̃,Jφ(xn+1-x̃)]. It follows that from condition (C1), limnxn+1-xn=0 and (3.47) that limnγn=0,n=1γn=,limsupnδn0. The inequality (3.57) reduces to the following: Φ(xn+1-x̃)(1-γn)Φ(xn-x̃)+γnδn. Applying Lemma 2.2, we conclude that Φ(xn+1-x̃)0 as n; that is, xnx̃ as n. This completes the proof.

Remark 3.7.

In comparison to the results in [13, Theorem 3.1], the strong convergence in a real Hilbert space is extended to the strong convergence in a reflexive Banach space which admits a weakly continuous duality mapping.

Setting BI, and μ1 in Theorem 3.6, we obtain the following result.

Corollary 3.8.

Let E be a reflexive Banach space which admits a weakly continuous duality mapping Jφ with gauge φ such that φ is invariant on [0,1]. Let T:EE be a nonexpansive mapping with F(T), f:EE a contraction with coefficient α(0,1) and A a strongly positive bounded linear operator with coefficient γ¯>0 and 0<γ<γ¯φ(1)/α. Let x0E be arbitrary, and let the sequence {xn} be generated by the following iterative scheme: xn+1=(I-λnA)Txn+λnγf(xn),n0, where {λn} is a real sequence in (0,1) satisfying the following conditions:

limnλn=0 and n=1λn=,

n=1|λn+1-λn|< or limn(λn/λn+1)=1.

Then, the sequence {xn} converges strongly to a fixed point x̃ of T which solves the variational inequality (A-γf)x̃,Jφ(x̃-z)0,zF(T).

Acknowledgments

The author is supported by Naresuan University and the Thailand Research Fund under Grant no. TRG5280011. Finally, the author would like to thank the referees for reading this paper carefully, providing valuable suggestions and comments and pointing out a major error in the original version of this paper.

BrowderF. E.Fixed-point theorems for noncompact mappings in Hilbert spaceProceedings of the National Academy of Sciences of the United States of America196553127212760178324ReichS.Strong convergence theorems for resolvents of accretive operators in Banach spacesJournal of Mathematical Analysis and Applications198075128729210.1016/0022-247X(80)90323-6576291XuH.-K.Strong convergence of an iterative method for nonexpansive and accretive operatorsJournal of Mathematical Analysis and Applications2006314263164310.1016/j.jmaa.2005.04.0822185255DeutschF.YamadaI.Minimizing certain convex functions over the intersection of the fixed point sets of nonexpansive mappingsNumerical Functional Analysis and Optimization1998191-2335610.1080/016305698088168131606921WangkeereeR.KamraksaU.A general iterative method for solving the variational inequality problem and fixed point problem of an infinite family of nonexpansive mappings in Hilbert spacesFixed Point Theory and Applications20092009233692152496029XuH.-K.An iterative approach to quadratic optimizationJournal of Optimization Theory and Applications2003116365967810.1023/A:10230736215891977756XuH.-K.Iterative algorithms for nonlinear operatorsJournal of the London Mathematical Society. Second Series200266124025610.1112/S00246107020033321911872MoudafiA.Viscosity approximation methods for fixed-points problemsJournal of Mathematical Analysis and Applications20002411465510.1006/jmaa.1999.66151738332PlubtiengS.ThammathiwatT.A viscosity approximation method for equilibrium problems, fixed point problems of nonexpansive mappings and a general system of variational inequalitiesJournal of Global Optimization201046344746410.1007/s10898-009-9448-52593241XuH.-K.Viscosity approximation methods for nonexpansive mappingsJournal of Mathematical Analysis and Applications2004298127929110.1016/j.jmaa.2004.04.0592086546MarinoG.XuH.-K.A general iterative method for nonexpansive mappings in Hilbert spacesJournal of Mathematical Analysis and Applications20063181435210.1016/j.jmaa.2005.05.0282210870WangkeereeR.PetrotN.WangkeereeR.The general iterative methods for nonexpansive mappings in Banach spacesJournal of Global Optimization. In press10.1007/s10898-010-9617-6CengL.-C.GuuS.-M.YaoJ.-C.Hybrid viscosity-like approximation methods for nonexpansive mappings in Hilbert spacesComputers & Mathematics with Applications200958360561710.1016/j.camwa.2009.02.0352538002TakahashiW.Nonlinear Functional Analysis. Fixed Point Theory and Its Applications2000Yokohama, JapanYokohama Publishersiv+2761864294BrowderF. E.Convergence theorems for sequences of nonlinear operators in Banach spacesMathematische Zeitschrift19671002012250215141LimT.-C.XuH. K.Fixed point theorems for asymptotically nonexpansive mappingsNonlinear Analysis: Theory, Methods & Applications199422111345135510.1016/0362-546X(94)90116-31280202