Volume 2009, Article ID 374815,32pages doi:10.1155/2009/374815
Research Article
A Hybrid Extragradient Viscosity
Approximation Method for Solving Equilibrium Problems and Fixed Point Problems of Infinitely Many Nonexpansive Mappings
Chaichana Jaiboon and Poom Kumam
Department of Mathematics, Faculty of Science, King Mongkut’s University of Technology Thonburi, Bangkok 10140, Thailand
Correspondence should be addressed to Poom Kumam,poom.kum@kmutt.ac.th Received 25 December 2008; Accepted 4 May 2009
Recommended by Wataru Takahashi
We introduce a new hybrid extragradient viscosity approximation method for finding the common element of the set of equilibrium problems, the set of solutions of fixed points of an infinitely many nonexpansive mappings, and the set of solutions of the variational inequality problems for β-inverse-strongly monotone mapping in Hilbert spaces. Then, we prove the strong convergence of the proposed iterative scheme to the unique solution of variational inequality, which is the optimality condition for a minimization problem. Results obtained in this paper improve the previously known results in this area.
Copyrightq2009 C. Jaiboon and P. Kumam. 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.
1. Introduction
LetH be a real Hilbert space, and let Cbe a nonempty closed convex subset ofH. Recall that a mappingT of H into itself is called nonexpansivesee1if Tx−Ty ≤ x−y for allx, y ∈H. We denote byFT {x ∈ C: Tx x}the set of fixed points ofT. Recall also that a self-mappingf :H → His a contraction if there exists a constantα∈0,1such thatfx−fy ≤αx−y, for all x, y ∈H.In addition, letB:C → Hbe a nonlinear mapping. LetPC be the projection ofHontoC. The classical variational inequality which is denoted byV IC, Bis to findu∈Csuch that
Bu, v−u ≥0, ∀v∈C. 1.1
For a givenz∈H,u∈Csatisfies the inequality
u−z, v−u ≥0, ∀v∈C, 1.2 if and only ifuPCz. It is well known thatPCis a nonexpansive mapping ofHontoCand satisfies
x−y, PCx−PCy
≥PCx−PCy2, ∀x, y∈H. 1.3 Moreover,PCxis characterized by the following properties:PCx∈Cand for allx∈H, y∈C,
x−PCx, y−PCx
≤0, 1.4
x−y2≥ x−PCx2y−PCx2. 1.5
It is easy to see that the following is true:
u∈V IC, B⇐⇒uPCu−λBu, λ >0. 1.6 One can see that the variational inequality1.1is equivalent to a fixed point problem.
The variational inequality has been extensively studied in literature; see, for instance, 2–
6. This alternative equivalent formulation has played a significant role in the studies of the variational inequalities and related optimization problems. Recall the following.
1A mappingBofCintoHis called monotone if Bx−By, x−y
≥0, ∀x, y∈C. 1.7
2A mappingB is calledβ-strongly monotonesee7,8if there exists a constant β >0 such that
Bx−By, x−y
≥βx−y2, ∀x, y∈C. 1.8 3A mappingBis calledk-Lipschitz continuous if there exists a positive real number
ksuch that
Bx−By≤kx−y, ∀x, y∈C. 1.9 4A mapping Bis called β-inverse-strongly monotonesee 7, 8 if there exists a
constantβ >0 such that Bx−By, x−y
≥βBx−By2, ∀x, y∈C. 1.10
Remark 1.1. It is obvious that anyβ-inverse-strongly monotone mappingBis monotone and 1/β-Lipschitz continuous.
5An operatorAis strongly positive onH if there exists a constantγ > 0 with the property
Ax, x ≥γx2, ∀x∈H. 1.11
6A set-valued mappingT :H → 2His called monotone if for allx, y∈H,f ∈Tx, andg ∈Tyimplyx−y, f−g ≥0. A monotone mappingT :H → 2His maximal if the graph ofGT ofT is not properly contained in the graph of any other monotone mapping. It is known that a monotone mappingTis maximal if and only if forx, f∈H×H,x−y, f−g ≥ 0 for everyy, g∈GTimpliesf∈Tx. LetBbe a monotone map ofCintoH,and letNCv be the normal cone toCatv∈C, that is,NCv{w∈H:u−v, w ≥0, for allu∈C},.
Tv
⎧⎨
⎩
BvNCv, v∈ C,
∅, v /∈ C. 1.12
ThenT is the maximal monotone and 0∈Tvif and only ifv∈V IC, B; see9.
7 Let F be a bifunction ofC×Cinto R, whereR is the set of real numbers. The equilibrium problem forF:C×C → Ris to findx∈Csuch that
F
x, y ≥0, ∀y∈C. 1.13
The set of solutions of 1.13 is denoted by EPF. Given a mapping T : C → H, let Fx, y Tx, y− x for all x, y ∈ C. Then, z ∈ EPF if and only if Tz, y−z ≥ 0 for all y ∈ C.Numerous problems in physics, saddle point problem, fixed point problem, variational inequality problems, optimization, and economics are reduced to find a solution of 1.13. Some methods have been proposed to solve the equilibrium problem; see, for instance, 10–16. Recently, Combettes and Hirstoaga 17introduced an iterative scheme of finding the best approximation to the initial data whenEPFis nonempty and proved a strong convergence theorem.
In 1976, Korpelevich18introduced the following so-called extragradient method:
x0x∈C,
ynPCxn−λBxn, xn1PC
xn−λByn
1.14
for alln≥0,whereλ∈0,1/k, Cis a closed convex subset ofRn,andBis a monotone and k-Lipschitz continuous mapping ofCintoRn. He proved that ifV IC, Bis nonempty, then the sequences{xn}and{yn}, generated by1.14, converge to the same pointz∈V IC, B.
For finding a common element of the set of fixed points of a nonexpansive mapping and
the set of solution of variational inequalities forβ-inverse-strongly monotone, Takahashi and Toyoda19introduced the following iterative scheme:
x0∈C chosen arbitrary,
xn1αnxn 1−αnSPCxn−λnBxn, ∀n≥0, 1.15 whereBisβ-inverse-strongly monotone,{αn}is a sequence in0, 1, and{λn}is a sequence in0,2β. They showed that ifFS∩V IC, Bis nonempty, then the sequence{xn}generated by1.15converges weakly to somez∈FS∩V IC, B. Recently, Iiduka and Takahashi20 proposed a new iterative scheme as follows:
x0x∈C chosen arbitrary,
xn1αnx 1−αnSPCxn−λnBxn, ∀n≥0, 1.16 whereBisβ-inverse-strongly monotone,{αn}is a sequence in0, 1, and{λn}is a sequence in0,2β. They showed that ifFS∩V IC, Bis nonempty, then the sequence{xn}generated by1.16converges strongly to somez∈FS∩V IC, B.
Iterative methods for nonexpansive mappings have recently been applied to solve convex minimization problems; see, for example,21–24and the references therein. Convex minimization problems have a great impact and influence in the development of almost all branches of pure and applied sciences. A typical problem is to minimize a quadratic function over the set of the fixed points of a nonexpansive mapping on a real Hilbert spaceH:
minx∈C
1
2Ax, x − x, b, 1.17
whereAis a linear bounded operator,Cis the fixed point set of a nonexpansive mapping SonH, andbis a given point inH. Moreover, it is shown in25that the sequence {xn} defined by the scheme
xn1 nγfxn 1−nASxn 1.18 converges strongly to z PFSI − A γfz. Recently, Plubtieng and Punpaeng 26 proposed the following iterative algorithm:
F
un, y 1 rn
y−un, un−xn
≥0, ∀y∈H, xn1nγfxn I−nASun.
1.19
They prove that if the sequences{n}and{rn}of parameters satisfy appropriate condition, then the sequences{xn}and{un}both converge to the unique solutionzof the variational inequality
A−γf q, q−p ≥0, p∈FS∩EPF, 1.20
which is the optimality condition for the minimization problem
x∈FS∩EPFmin 1
2Ax, x −hx, 1.21 wherehis a potential function forγfi.e.,hx γfxforx∈H.
Furthermore, for finding approximate common fixed points of an infinite countable family of nonexpansive mappings {Tn} under very mild conditions on the parameters.
Wangkeeree27introduced an iterative scheme for finding a common element of the set of solutions of the equilibrium problem1.13and the set of common fixed points of a countable family of nonexpansive mappings onC. Starting with an arbitrary initialx1 ∈ C, define a sequence{xn}recursively by
F
un, y 1 rn
y−un, un−xn
≥0, ∀y∈C, ynPCun−λnBun,
xn1 αnfxn βnxnγnSnPC
un−λnByn , ∀n≥1,
1.22
where{αn},{βn},and{γn}are sequences in0,1. It is proved that under certain appropriate conditions imposed on {αn},{βn},{γn}, and {rn}, the sequence {xn} generated by 1.22 strongly converges to the unique solution q ∈ ∩∞n1FSn∩V IC, B∩EPF, where p P∩∞n1FSn∩V IC,B∩EPFfqwhich extend and improve the result of Kumam14.
Definition 1.2see21. Let{Tn}be a sequence of nonexpansive mappings ofCinto itself, and let{μn}be a sequence of nonnegative numbers in0,1. For eachn≥1, define a mapping WnofCinto itself as follows:
Un,n1I,
Un,nμnTnUn,n1
1−μn I, Un,n−1μn−1Tn−1Un,n
1−μn−1 I, ...
Un,kμkTkUn,k1
1−μk I, Un,k−1μk−1Tk−1Un,k
1−μk−1 I, ...
Un,2μ2T2Un,3
1−μ2 I, WnUn,1μ1T1Un,2
1−μ1 I.
1.23
Such a mappingWnis nonexpansive fromCtoC,and it is called theW-mapping generated byT1, T2, . . . , Tnandμ1, μ2, . . . , μn.
On the other hand, Colao et al.28introduced and considered an iterative scheme for finding a common element of the set of solutions of the equilibrium problem1.13and the set of common fixed points of infinitely many nonexpansive mappings onC. Starting with an arbitrary initialx0∈C, define a sequence{xn}recursively by
F
un, y 1
rny−un, un−xn ≥0, ∀y∈H, xn1nγfxn βxn
1−β I−nA Wnun,
1.24
where{n}is a sequence in0,1. It is proved28that under certain appropriate conditions imposed on{n}and{rn}, the sequence{xn}generated by1.24strongly converges toz ∈
∩∞n1FTn∩EPF, wherezis an equilibrium point forF and is the unique solution of the variational inequality1.20, that is,zP∩∞n1FTn∩EPFI−A−γfz.
In this paper, motivated by Wangkeeree27, Plubtieng and Punpaeng26, Marino and Xu25, and Colao, et al.28, we introduce a new iterative scheme in a Hilbert spaceH which is mixed by the iterative schemes of1.18,1.19,1.22, and1.24as follows.
Letfbe a contraction ofHinto itself,Aa strongly positive bounded linear operator on Hwith coefficientγ >0,andBaβ-inverse-strongly monotone mapping ofCintoH; define sequences{xn},{yn},{kn},and{un}recursively by
x1 x∈C chosen arbitrary, F
un, y 1
rny−un, un−xn ≥0, ∀y∈C, ynPCun−λnBun,
knαnun 1−αnPC
un−λnByn , xn1nγfxn βnxn
1−βn I−nA Wnkn, ∀n≥1,
1.25
where {Wn}is the sequence generated by 1.23,{n},{αn},and {βn} ⊂ 0,1 and {rn} ⊂ 0,∞satisfying appropriate conditions. We prove that the sequences{xn},{yn},{kn}and {un}generated by the above iterative scheme1.25converge strongly to a common element of the set of solutions of the equilibrium problem1.13, the set of common fixed points of infinitely family nonexpansive mappings, and the set of solutions of variational inequality 1.1for a β-inverse-strongly monotone mapping in Hilbert spaces. The results obtained in this paper improve and extend the recent ones announced by Wangkeeree 27, Plubtieng and Punpaeng26, Marino and Xu25, Colao, et al.28, and many others.
2. Preliminaries
We now recall some well-known concepts and results.
LetHbe a real Hilbert space, whose inner product and norm are denoted by·,·and · , respectively. We denote weak convergence and strong convergence by notationsand
→, respectively.
A spaceHis said to satisfy Opial’s condition29if for each sequence{xn}inHwhich converges weakly to pointx∈H, we have
lim inf
n→ ∞ xn−x<lim inf
n→ ∞ xn−y, ∀y∈H, y /x. 2.1 Lemma 2.1see25. LetCbe a nonempty closed convex subset ofH, letf be a contraction of H into itself withα ∈ 0,1, and letAbe a strongly positive linear bounded operator on Hwith coefficientγ >0. Then , for 0< γ < γ/α,
x−y,
A−γf x−
A−γf y
≥
γ−αγ x−y2, x, y∈H. 2.2 That is,A−γfis strongly monotone with coefficientγ−γα.
Lemma 2.2see25. Assume thatA is a strongly positive linear bounded operator onH with coefficientγ >0 and 0< ρ≤ A−1. ThenI−ρA ≤1−ργ.
For solving the equilibrium problem for a bifunctionF :C×C → R, let us assume thatFsatisfies the following conditions:
A1Fx, x 0 for allx∈C;
A2Fis monotone, that is,Fx, y Fy, x≤0 for allx, y∈C;
A3for eachx, y, z∈C,limt↓0Ftz 1−tx, y≤Fx, y;
A4for eachx∈C, y→Fx, yis convex and lower semicontinuous.
The following lemma appears implicitly in30.
Lemma 2.3see30. LetCbe a nonempty closed convex subset ofHand letFbe a bifunction of C×CintoRsatisfying (A1)–(A4). Letr >0 andx∈H. Then, there existsz∈Csuch that
F
z, y 1 r
y−z, z−x
≥0 ∀y∈C. 2.3
The following lemma was also given in17.
Lemma 2.4see17. Assume thatF :C×C → Rsatisfies (A1)–(A4). Forr > 0 andx ∈H, define a mappingTr :H → Cas follows:
Trx
z∈C:F
z, y 1 r
y−z, z−x
≥0, ∀y∈C
2.4
for allz∈H. Then, the following holds:
1Tr is single-valued;
2Tr is firmly nonexpansive, that is, for anyx, y∈H, Trx−Try2≤
Trx−Try, x−y
; 2.5
3FTr EPF;
4EPFis closed and convex.
For eachn, k ∈ N, let the mappingUn,kbe defined by1.23. Then we can have the following crucial conclusions concerningWn. You can find them in31. Now we only need the following similar version in Hilbert spaces.
Lemma 2.5see31. LetC be a nonempty closed convex subset of a real Hilbert spaceH. Let T1, T2, . . . be nonexpansive mappings of C into itself such that ∩∞n1FTn is nonempty, and let μ1, μ2, . . .be real numbers such that 0 ≤ μn ≤ b < 1 for every n ≥ 1. Then, for everyx ∈ Cand k∈N, the limit limn→ ∞Un,kxexists.
UsingLemma 2.5, one can define a mappingWofCinto itself as follows:
Wx lim
n→ ∞Wnx lim
n→ ∞Un,1x 2.6
for everyx ∈ C. Such aW is called the W-mapping generated byT1, T2, . . . and μ1, μ2, . . ..
Throughout this paper, we will assume that 0≤μn≤b <1 for everyn≥1. Then, we have the following results.
Lemma 2.6see31. LetC be a nonempty closed convex subset of a real Hilbert spaceH. Let T1, T2, . . . be nonexpansive mappings of C into itself such that ∩∞n1FTn is nonempty, and let μ1, μ2, . . .be real numbers such that 0≤μn≤b <1 for everyn≥1. Then,FW ∩∞n1FTn. Lemma 2.7see32. If{xn}is a bounded sequence inC, then limn→ ∞Wxn−Wnxn0.
Lemma 2.8see33. Let{xn}and{zn}be bounded sequences in a Banach spaceX,and let{βn}be a sequence in0,1with 0<lim infn→ ∞βn≤lim supn→ ∞βn<1.Supposexn1 1−βnznβnxn
for all integersn≥0 and lim supn→ ∞yn1−zn − xn1−xn≤0.Then, limn→ ∞zn−xn0.
Lemma 2.9see34. Assume that{an}is a sequence of nonnegative real numbers such that
an1≤1−lnanσn, n≥0, 2.7
where{ln}is a sequence in0,1and{σn}is a sequence inRsuch that 1∞
n1ln∞;
2lim supn→ ∞σn/ln≤0 or∞
n1|σn|<∞.
Then limn→ ∞an0.
Lemma 2.10. LetHbe a real Hilbert space. Then for allx, y∈H, 1xy2≤ x22y, xy;
2xy2≥ x22y, x.
3. Main Results
In this section, we prove the strong convergence theorem for infinitely many nonexpansive mappings in a real Hilbert space.
Theorem 3.1. LetCbe a nonempty closed convex subset of a real Hilbert spaceH, letFbe a bifunction fromC×CtoRsatisfying (A1)–(A4), let{Tn}be an infinitely many nonexpansive ofCinto itself, and let B be an β-inverse-strongly monotone mapping ofC into H such that Θ : ∩∞n1FTn∩ EPF∩V IC, B/∅. Letfbe a contraction ofHinto itself withα∈0,1,and letAbe a strongly positive linear bounded operator onH with coefficient γ > 0 and 0 < γ < γ/α. Let {xn},{yn}, {kn},and{un}be sequences generated by1.25, where{Wn}is the sequence generated by1.23, {n},{αn},and{βn}are three sequences in0,1,and{rn}is a real sequence in0,∞satisfying the following conditions:
ilimn→ ∞n0,∞
n1n∞;
iilimn→ ∞αn0 and∞
n1αn∞;
iii0<lim infn→ ∞βn≤lim supn→ ∞βn<1;
ivlim infn→ ∞rn>0 and limn→ ∞|rn1−rn|0;
v{λn/β} ⊂τ,1−δfor someτ, δ∈0,1and limn→ ∞λn0.
Then,{xn}and{un}converge strongly to a pointz∈Θwhich is the unique solution of the variational inequality
A−γf z, z−x
≥0, ∀x∈Θ. 3.1
Equivalently, one haszPΘI−Aγfz.
Proof. Note that from the conditioni, we may assume, without loss of generality, thatn ≤ 1−βnA−1for alln∈N. FromLemma 2.2, we know that if 0≤ρ≤ A−1, thenI−ρA ≤ 1−ργ. We will assume thatI−A ≤1−γ. First, we show thatI−λnBis nonexpansive. Indeed, from theβ-inverse-strongly monotone mapping definition onBand conditionv, we have
I−λnBx−I−λnBy2x−y−λnBx−By2 x−y2−2λn
x−y, Bx−By
λ2nBx−By2
≤x−y2−2λnβBx−By2λ2nBx−By2 x−y2λn
λn−2β Bx−By2
≤x−y2,
3.2
which implies that the mapping I−λnB is nonexpansive. On the other hand, sinceAis a strongly positive bounded linear operator on H, we have
Asup{|Ax, x|:x∈H,x1}. 3.3
Observe that
1−βn I−nA x, x
1−βn−nAx, x
≥1−βn−nA
≥0,
3.4
and this show that1−βnI−nAis positive. It follows that
1−βn I−nAsup1−βn I−nA x, x:x∈H,x1 sup
1−βn−nAx, x:x∈H,x1
≤1−βn−nγ.
3.5
LetQPΘ, whereΘ:∩∞n1FTn∩EPF∩V IC, B. Note thatf is a contraction ofHinto itself withα∈0,1. Then, we have
Q
I−Aγf x−Q
I−Aγf y PΘ
I−Aγf x−PΘ
I−Aγf y
≤I−Aγf x−
I−Aγf y
≤ I−Ax−yγfx−f y
≤
1−γ x−yγαx−y
1−γγα x−y
1−
γ−γα x−y, ∀x, y∈H.
3.6
Since 0 < 1−γ −γα < 1, it follows that QI−Aγf is a contraction ofH into itself.
Therefore by the Banach Contraction Mapping Principle, which implies that there exists a unique elementz∈Hsuch thatzQI−Aγfz PΘI−Aγfz.
We will divide the proof into five steps.
Step 1. We claim that{xn}is bounded. Indeed, pick anyp∈Θ. From the definition ofTr, we note thatunTrnxn. If follows that
un−pTrnxn−Trnp≤xn−p. 3.7 SinceI−λnBis nonexpansive andpPCp−λnBpfrom1.6, we have
yn−pPCun−λnBun−PC
p−λnBp
≤un−λnAun−
p−λnBp I−λnAun−I−λnBp
≤un−p≤xn−p.
3.8
Putvn PCun−λnByn. Sincep ∈ V IC, B, we havep PCp−λnBp. Substitutingx un−λnAynandypin1.5, we can write
vn−p2 ≤un−λnByn−p2−un−λnByn−vn2 un−p2−2λn
Byn, un−p
λ2nByn2
− un−vn22λn
Byn, un−vn
−λ2nByn2 un−p2− un−vn22λn
Byn, p−vn un−p2− un−vn22λn
Byn−Bp, p−yn 2λn
Bp, p−yn 2λn
Byn, yn−vn .
3.9
Using the fact that B is β-inverse-strongly monotone mapping, and p is a solution of the variational inequality problemV IC, B, we also have
Byn−Bp, p−yn
≤0, Bp, p−yn ≤0. 3.10
It follows from3.9and3.10that
vn−p2≤un−p2− un−vn22λnByn, yn−vn un−p2−un−yn yn−vn22λn
Byn, yn−vn
≤un−p2−un−yn2−yn−vn2
−2
un−yn, yn−vn 2λn
Byn, yn−vn un−p2−un−yn2−yn−vn22
un−λnByn−yn, vn−yn .
3.11
Substitutingxbyun−λnBunandyvnin1.4, we obtain un−λnBun−yn, vn−yn
≤0. 3.12
It follows that
un−λnByn−yn, vn−yn
un−λnBun−yn, vn−yn
λnBun−λnByn, vn−yn
≤
λnBun−λnByn, vn−yn
≤λnBun−Bynvn−yn
≤ λn
β un−ynvn−yn.
3.13
Substituting3.13into3.11, we have
vn−p2≤un−p2−un−yn2−yn−vn22
un−λnByn−yn, vn−yn
≤un−p2−un−yn2−yn−vn22λn
β un−ynvn−yn
≤un−p2−un−yn2−yn−vn2λ2n
β2un−yn2vn−yn2 un−p2−un−yn2λ2n
β2un−yn2 un−p2
λ2n β2 −1
un−yn2
≤un−p2≤xn−p2.
3.14
Settingknαnun 1−αnvn, we can calculate xn1−pn
γfxn−Ap βn
xn−p
1−βn I−nA Wnkn−p
≤
1−βn−nγ kn−pβnxn−pnγfxn−Ap
≤
1−βn−nγ αnun−p 1−αnvn−p βnxn−pnγfxn−Ap
≤
1−βn−nγ αnxn−p 1−αnxn−p βnxn−pnγfxn−Ap
1−βn−nγ xn−pβnxn−pnγfxn−Ap
1−nγ xn−pnγfxn−f
p nγf
p −Ap
≤
1−nγ xn−pnγαxn−pnγf
p −Ap
1−
γ−γα n xn−p
γ−γα n
γf
p −Ap γ−γα .
3.15
By induction,
xn−p≤max
x1−p,γf
p −Ap γ−γα
, n∈N. 3.16
Hence,{xn}is bounded, so are{un},{vn},{Wnkn},{fxn},{Bun},{yn},and{Byn}.
Step 2. We claim that limn→ ∞xn1−xn0.
Observing thatunTrnxnandun1Trn1xn1,we get
F
un, y 1 rn
y−un, un−xn
≥0 ∀y∈H 3.17 F
un1, y 1 rn1
y−un1, un1−xn1
≥0 ∀y∈H. 3.18
Puttingyun1in3.17andyunin3.18, we have
Fun, un1 1
rnun1−un, un−xn ≥0 Fun1, un 1
rn1un−un1, un1−xn1 ≥0.
3.19
So, fromA2we have
un1−un,un−xn
rn −un1−xn1 rn1
≥0, 3.20
and hence
un1−un, un−un1un1−xn− rn
rn1un1−xn1
≥0. 3.21
Without loss of generality, let us assume that there exists a real numbercsuch thatrn> c >0 for alln∈N.Then, we have
un1−un2≤
un1−un, xn1−xn
1− rn rn1
un1−xn1
≤ un1−un
xn1−xn 1− rn
rn1
un1−xn1
,
3.22
and hence
un1−un ≤ xn1−xn 1
rn1|rn1−rn|un1−xn1
≤ xn1−xnM1
c |rn1−rn|,
3.23
whereM1sup{un−xn:n∈N}. Note that vn1−vn ≤PC
un1−λn1Byn1 −PC
un−λnByn
≤un1−λn1Byn1−
un−λnByn un1−λn1Bun1−un−λn1Bun
λn1
Bun1−Byn1−Bun λnByn
≤ un1−λn1Bun1−un−λn1Bun λn1
Bun1Byn1Bun λnByn
≤ un1−unλn1
Bun1Byn1Bun λnByn, kn1−knαn1un1 1−αn1vn1−αnun−1−αnvn
αn1un1−un αn1−αnun
1−αn1vn1−vn αn−αn1vn
≤αn1un1−un 1−αn1vn1−vn|αn−αn1|unvn αn1un1−un 1−αn1
×
un1−unλn1
Bun1Byn1Bun λnByn
|αn−αn1|unvn un1−un 1−αn1λn1
Bun1Byn1Bun 1−αn1λnByn|αn−αn1|unvn
≤ xn1−xnM1
c |rn1−rn| 1−αn1λn1
×
Bun1Byn1Bun
1−αn1λnByn|αn−αn1|unvn.
3.24
Setting
zn xn1−βnxn
1−βn nγfxn
1−βn I−nA Wnkn
1−βn , 3.25
we havexn1 1−βnznβnxn, n≥1. It follows that zn1−zn n1γfxn1
1−βn1 I−n1A Wn1kn1 1−βn1
− nγfxn
1−βn I−nA Wnkn 1−βn
n1
1−βn1γfxn1− n
1−βnγfxn Wn1kn1−Wnkn n
1−βnAWnkn− n1
1−βn1AWn1kn1 n1
1−βn1
γfxn1−AWn1kn1 n 1−βn
AWnkn−γfxn
Wn1kn1−Wn1knWn1kn−Wnkn.
3.26
It follows from3.24and3.26that zn1−zn − xn1−xn ≤ n1
1−βn1γfxn1AWn1kn1 n
1−βn
AWnknγfxn Wn1kn1−Wn1kn
Wn1kn−Wnkn − xn1−xn
≤ n1
1−βn1γfxn1AWn1kn1 n
1−βn
AWnknγfxn kn1−kn
Wn1kn−Wnkn − xn1−xn
≤ n1
1−βn1γfxn1AWn1kn1 n
1−βn
AWnknγfxn M1
c |rn1−rn| 1−αn1λn1
Bun1Byn1Bun 1−αn1λnByn|αn−αn1|unvn Wn1kn−Wnkn.
3.27
SinceTiandUn,iare nonexpansive, we have
Wn1kn−Wnknμ1T1Un1,2kn−μ1T1Un,2kn
≤μ1Un1,2kn−Un,2kn
μ1μ2T2Un1,3kn−μ2T2Un,3kn
≤μ1μ2Un1,3kn−Un,3kn ...
≤μ1μ2· · ·μnUn1,n1kn−Un,n1kn
≤M2n
i1
μi,
3.28
whereM2≥0 is a constant such thatUn1,n1kn−Un,n1kn ≤M2for alln≥0.
Combining3.27and3.28, we have zn1−zn − xn1−xn ≤ n1
1−βn1γfxn1AWn1kn1 n
1−βn
AWnknγfxn M1
c |rn1−rn| 1−αn1λn1
Bun1Byn1Bun 1−αn1λnByn|αn−αn1|unvn M2n
i1
μi,
3.29
which implies thatnoting thati,ii,iii,iv,v, and 0< μi≤b <1,for alli≥1 lim sup
n→ ∞ zn1−zn − xn1−xn≤0. 3.30
Hence, byLemma 2.8, we obtain
n→ ∞limzn−xn0. 3.31
It follows that
nlim→ ∞xn1−xn lim
n→ ∞
1−βn zn−xn0. 3.32
Applying3.32andii,iv, andvto3.23and3.24, we obtain that
nlim→ ∞un1−un lim
n→ ∞kn1−kn0. 3.33
Sincexn1nγfxn βnxn 1−βnI−nAWnkn, we have xn−Wnkn ≤ xn−xn1xn1−Wnkn
xn−xn1nγfxn βnxn
1−βn I−nA Wnkn−Wnkn xn−xn1n
γfxn−AWnkn βnxn−Wnkn
≤ xn−xn1nγfxnAWnkn βnxn−Wnkn,
3.34
that is
xn−Wnkn ≤ 1
1−βnxn−xn1 n
1−βn
γfxnAWnkn . 3.35
Byi,iii, and3.32it follows that
n→ ∞limWnkn−xn0. 3.36
Step 3. We claim that the following statements hold:
ilimn→ ∞un−kn0;
iilimn→ ∞xn−un0.
For anyp∈Θ:∩∞n1FTn∩EPF∩V IC, Band3.14, we have kn−p2αnun−p 1−αnvn−p2
≤αnun−p2 1−αnvn−p2
≤αnun−p2 1−αn
un−p2 λ2n
β2 −1
un−yn2 un−p2 1−αn
λ2n β2 −1
un−yn2
≤xn−p2 1−αn λ2n
β2 −1
un−yn2.
3.37