vjom-d-13-00067r2

Upload: khuongnn

Post on 03-Jun-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 VJOM-D-13-00067R2

    1/19

    Vietnam Journal of Mathematics

    (will be inserted by the editor)

    A Fixed Point Scheme for Nonexpansive Mappings, Variational

    Inequalities and Equilibrium Problems

    Pham N. Anh Le Q. Thuy Do D. Thanh

    Received: 10 May 2013 / Accepted: 15 December 2013

    Abstract The purpose of this paper is to introduce a new iteration scheme and prove a

    strong convergence theorem for finding the common element of the fixed point set of a

    nonexpansive mapping, the solution set of variational inequalities and the solution set of

    equilibrium problems. Under certain conditions on parameters, we show that the iterative

    sequences generated by the scheme strongly converge to the common element in a real

    Hilbert space.

    Keywords Nonexpansive Pseudomonotone Continuous Fixed point Variationalinequalities Equilibrium problems

    Mathematics Subject Classification (2010) 65K10 90C33

    1 Introduction

    The equilibrium problems in the sense of Blum and Oettli (see [7]), shortly EP (f,C), arepresented as follows:

    Find x C such that f(x,y) 0 for ally C, EP(f,C)

    whereCis a nonempty closed convex subset of a real Hilbert space Hwith inner product

    , and norm , f :C C R is a bifunction such that f(x,x) = 0 for all x Cand

    This work is supported by the Vietnam Institute for Advanced Study in Mathematics.

    P.N. Anh ( )Department of Scientific Fundamentals, Posts and Telecommunications Institute of Technology, Hanoi, Viet-

    nam

    E-mail: [email protected]

    L.Q. Thuy

    School of Applied Mathematics and Informatics, Ha Noi University of Science and Technology, No. 1, Dai

    Co Viet Road, Hai Ba Trung, Hanoi, Vietnam

    E-mail: [email protected]

    D.D. Thanh

    Department of Mathematics, Haiphong University, Vietnam

  • 8/12/2019 VJOM-D-13-00067R2

    2/19

    2 Pham N. Anh et al.

    f(x, ) is convex and subdifferentiable on Cfor every x C. The solution set of ProblemEP(f,C) is denoted by Sol(f,C). The bifunction f is called strongly monotone onCwith> 0, if

    f(x,y) +f(y,x) x y2 x,y C;

    monotone on C, if

    f(x,y) +f(y,x) 0 x,y C;

    pseudomonotoneonC, if

    f(x,y) 0 f(y,x) 0 x,y C;

    Lipschitz-type continuouson Cwith constantsc1> 0 and c2> 0 in the sense of Mastroeni(see [13]) if

    f(x,y) +f(y,z) f(x,z) c1x y2 c2y z

    2 x,y,z C.

    The mappingB : C H is called-inverse strongly monotone, if

    Bx By,x y Bx By2 x,y C.

    The variational inequalities (shortly VI(C,B)) are to findx Csuch that

    Bx,x x 0 x C. (1)

    The solution set of Problem VI(C,B)is denoted by sol(C,B). The mappingg : C Cis saidto becontractivewith (0,1), if

    g(x) g(y) x y x,y C.

    The mappingS: C Cis called nonexpansive, if

    Sx Sy x y x,y C.

    Fix(S)denotes the set of fixed points ofS. A linear bounded operator A ofH into itself iscalledstrongly positiveif there is a constant> 0 with property that

    Ax,x x2 x H.

    In this paper, we are interested in the problem of finding a common element of the

    solution set of the equilibrium problems Sol(f,C), the solution set of variational inequalitiessol(C,B)and the set of fixed points Fix(S), namely:

    Find x Fix(S) sol(C,B) Sol(f,C). (2)

    Problem (2) is very general and it includes the variational inequalities, the equilibrium prob-

    lems, the fixed point problem, and the vector optimization problem as particular cases. It

    is well-known that the problem of finding a common fixed point of mappings has been ex-

    tensively investigated (see, for example, [3,6,12] and the references therein). In fact, x is

    a solution to Problem VI(F,C) if and only if it is a fixed point of the mapping T, whereT(x):=PrC(x F(x))for allx C. Otherwise, for each> 0 andx C,x

    Sol(f,C)if and only if it is a fixed point of the solution mapping S, whereS(x):=argmin{f(x,y) +x y2 : y C} (see [1]). So, we can say that, solving the problem of finding a commonelement of the solution set of the equilibrium problem EP(f,C), the variational inequality

  • 8/12/2019 VJOM-D-13-00067R2

    3/19

    A Fixed Point Scheme for Nonexpansive Mappings 3

    problem VI(F,C) and the set of fixed points of nonexpansive mappings, is an extended studyof the problem of finding a common fixed point of mappings.

    Theory and methods for solving Problem (2) have been well developed by many re-

    searchers (see [2, 46,8,14, 18,19] and the references therein). Combettes and Hirstoaga

    [9] proposed an iterative scheme of finding the best approximation to the initial data when

    Sol(f,C) = /0 and proved a strong convergence theorem. Motivated by this idea, Takahashiand Takahashi [16] introduced the viscosity approximation method for finding a common

    element of the set of solutions of an equilibrium problem and the set of fixed points of a

    nonexpansive mapping in a real Hilbert space. Recently, Anh [1] proposed a hybrid extra-

    gradient algorithm for finding a point of the set Fix(S) Sol(f,C). The iteration sequence isgiven by

    x0 C,

    yk =argmin{kf(xk,y) + y xk2 :y C},

    tk =argmin{kf(yk,y) + y xk2 :y C},

    xk+1 = kx0 + (1 k)Sx

    k.

    (3)

    Under certain conditions on parameters kandk, he showed that the sequences {xk} and

    {yk} weakly converge to the point x Fix(S) Sol(f,C). Inspired and motivated by the ideaof Anh [1], Wangkeeree and Preechasilp [17] replaced xk+1 in (3) by

    xk+1 :=PC

    kg(x

    k) + (1 kA)SPC(tkBtk)

    ,

    wheretk =argmin{kf(y

    k,y) + y xk2 :y C}

    andg is contractive,A is a strongly positive linear bounded operator ofH into itself. Then

    they showed that under some control conditions, the sequences {xk},{yk}, and {tk} stronglyconverge to an element of Fix(S) sol(B,C) Sol(f,C). Combining the iteration method in[17] and an improvement set of linesearch techniques for equilibrium problems, we intro-

    duce a new iteration scheme for finding an element of Fix(S) sol(C,B) Sol(f,C). Then,we show that the iterative sequences generated by the scheme strongly converge to a solution

    of Problem (2) in a real Hilbert space.

    2 Preliminaries

    LetCbe a nonempty closed convex subset of a real Hilbert space H. Let f :C C Rbe a bifunction. For solving the mixed equilibrium problem, let us assume the following

    conditions for a bifunction f:

    (A1) f(x,x) =0 for all x C;(A2) fis continuous onC;(A3) fis pseudomonotone onC;(A4) for eachx C, f(x, )is convex and subdifferentiable on C.

    For each pointx H, there exists the unique nearest point in C, denoted byPC(x), such that

    x PC(x) x y y C.

    The mappingPCis called the metric projection on C.

  • 8/12/2019 VJOM-D-13-00067R2

    4/19

    4 Pham N. Anh et al.

    Lemma 1 Let C be a nonempty closed convex subset ofH, x H and y C. Then

    (i) y=PC(x)if and only ifx y,y z 0for all z C;(ii) PCis a nonexpansive mapping on C;

    (iii) x y,PC(x) PC(y) PC(x) PC(y)2 for all x,y H,

    (iv) x PC(x),PC(x) y 0for all x H and y C.

    Using Lemma 1, one can show that Problem (1) is equivalent to a fixed point problem.

    Lemma 2 A point u C is a solution of the variational inequality(1) if and only if u satisfiesthe relation u=P

    C(u Bu)for all> 0.

    A set-valued mappingT : H 2H is called monotone if for all x,y H,u T x, andv Ty imply x y,u v 0. A monotone mappingT: H 2H ismaximalif the graphG(T) ofTis not properly contained in the graph of any other monotone mapping. It is knownthat a monotone mappingTis maximal if and only if for (x,u) H H, x y,u v 0for every(y,v) G(T) impliesu T x. LetB be an inverse-strongly monotone mapping ofCto H, letNC(v)be the normal cone to Catv C, that is,NC(v) = {w H : v u,w 0 for all u C}, and define

    T v=

    Bv +NC(v) if v C,

    /0 if v/ C.(4)

    ThenTis a maximal monotone and v T10 if and only ifv sol(C,B)(see [15]).

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

    Lemma 3 (see [10]) Let C be a nonempty closed convex subset of a real Hilbert space H

    and h: C R be convex and subdifferentiable on C. Then x is a solution to the followingconvex problem:

    min{h(x): x C}

    if and only if0 h(x) +NC(x), where NC(x

    )is the out normal cone at x to C andh()denotes the subdifferential of g.

    Lemma 4 (see [11]) Let C be a closed convex subset of a real Hilbert space H and let

    S: C C be a nonexpansive mapping such thatFix(S) = /0. If a sequence {xn} C is suchthat xn z and xn Sxn 0, then z=Sz.

    Lemma 5 (see [12]) Assume that A is a strongly positive linear bounded operator on a

    Hilbert space Hwith coefficient> 0and0

  • 8/12/2019 VJOM-D-13-00067R2

    5/19

    A Fixed Point Scheme for Nonexpansive Mappings 5

    Algorithm 1 Choose positive sequences {n} (0,1); [a,b],0

  • 8/12/2019 VJOM-D-13-00067R2

    6/19

    6 Pham N. Anh et al.

    Proof(i) For anyx,y H, we have

    P(IA +g)(x) P(IA +g)(y) (IA + g)(x) (IA + g)(y)

    g(x) g(y) + IAx y

    x y + (1 )x y

    =

    1 ( )

    x y.

    The Banachs contraction principle guarantees that P(IA +g)has a unique fixed point,saysq H. That is,q=P(IA +g)(q). By Lemma 1(i), we obtain

    (g A)q,x q 0 x . (9)

    (ii) If there existsn0 such thatd(xn) = 0, i.e.,xn =yn for alln n0, from Step 1, we have

    tn =xn. By Rabian Wangkeeree and Pakkapon Preechasilp [17], the sequences {xn},{yn},and {tn} strongly converge to the same point q .

    Now we considerd(xn) =0 and divide the proof into several steps.Step 1.Ifd(xn) =0, then there exists the smallest nonnegative integer mnsuch that

    f(xn mn d(xn),yn) d(xn)2.

    Indeed, for d(xn) =0 and (0,1), we suppose for contradiction that for everynonnegative integerm, we have

    fxn md(xn),yn+d(xn)2 >0.Passing to the limit of the above inequality as m , by continuity of f, we obtain

    fxn,yn

    +d(xn)2 0. (10)

    On the other hand, sinceyn is the unique solution of the strongly convex problem

    min

    nf(x

    n,y) +1

    2y xn2 : y C

    ,

    we have

    nf(xn,y) +

    1

    2y xn2 nf(x

    n,yn) +1

    2yn xn2 y C.

    Withy=xn, the last inequality implies

    nf(xn,yn) +12

    d(xn)||2 0. (11)

    Combining (10) with (11), we obtain

    d(xn)2 1

    2nd(xn)2.

    Hence it must be either d(xn)= 0 or 12n

    . The first case contradicts d(xn) =0,

    while the second one contradicts the fact n 12>1.

    Step 2.We claim that ifd(xn) =0 thenxn /Hn.Indeed, fromzn =xn mn d(xn), it follows that

    yn zn =1 mn

    mn(zn xn).

  • 8/12/2019 VJOM-D-13-00067R2

    7/19

  • 8/12/2019 VJOM-D-13-00067R2

    8/19

    8 Pham N. Anh et al.

    Indeed, suppose x , this implies that x Sol(f,C). Using the definition of x,f(x,x) 0 for allx Cand fis pseudomonotone onC, we get

    f(zn,x) 0. (14)

    It follows fromvn 2f(zn,zn)that

    f(zn,x) = f(zn,x) f(zn,zn)

    vn,x zn. (15)

    Combining (14) and (15), we have

    vn,x zn 0.

    By the definition ofHn, we havex Hn. Thus Sol(f,C) CHn, which means CHn.

    Step 5.Letwn = PC(tn Btn)andx . We claim that ifd(xn) = 0 then the sequences

    {xn},{wn} and {tn} are bounded.

    Indeed, in the case d(xn) =0, by Step 3, we have tn =PCHn ( yn), i.e.,

    yn tn,z tn 0 z CHn,

    where yn =PHn (xn). Substitutingz=x Sol(f,C) CHnby Step 4, then we have

    yn tn,x tn 0 yn tn,x yn +yn tn 0,

    which implies that

    tn yn2 tn yn,x yn.

    Hence

    tn x2 =tn yn +yn x2

    =tn yn2 + yn x2 + 2tn yn,yn x

    x yn, tn yn + yn x2 + 2tn yn,yn x

    = yn x2 + tn yn,yn x

    yn x2 tn yn2. (16)

    Sincezn =xn mn d(xn)and

    yn =PHn (xn) =xn

    vn,xn zn

    vn2 vn,

  • 8/12/2019 VJOM-D-13-00067R2

    9/19

    A Fixed Point Scheme for Nonexpansive Mappings 9

    we have

    yn x2 =xn x2 +vn,xn zn2

    vn4 vn2

    2vn,xn zn

    vn2 vn,xn x

    =xn x2 +

    mn vn,d(xn)

    vn

    2

    2mn vn,d(xn)

    vn2 vn,xn x

    =xn x2

    mn vn,d(xn)

    vn

    2

    2mn vn,d(xn)

    vn2 vn,xn x mn vn,d(xn)

    vn2

    =xn x2

    mn vn,d(xn)

    vn

    2

    2mn vn,d(xn)

    vn2 (vn,xn x mn vn,d(xn))

    =xn x2

    mn vn,d(xn)

    vn

    2

    2mn vn,d(xn)

    vn2 vn,xn x mn d(xn)

    =xn x2

    mn vn,d(xn)

    vn

    2

    2mn vn,d(xn)

    vn2 vn,zn x. (17)

    It follows fromvn 2f(zn,zn)that

    f(zn,y) f(zn,zn) vn,y zn y C. (18)

    Replacingy by yn and combining with assumptions f(zn,zn) = 0 and zn =xn mn d(xn),we have

    f(zn,yn) vn,yn zn

    =(1 mn )vn,d(xn).

    Combining this inequality with (6) and assumption (0,1), we obtain

    vn,d(xn)

    1 mnd(xn)2. (19)

    Substitutingy=x in (18) and using f(zn,zn) =0, we have

    f(zn,x) vn,x zn. (20)

    Since fis pseudomonotone onCand f(x,x) 0 x C, we have

    f(zn,x) 0.

    Combining this with (20), we get

    0 vn,x zn. (21)

    Using (17), (19) and (21), we have

    yn x2 xn x2

    mn vn,d(xn)

    vn

    2

    xn x2

    mn

    vn(1 mn )2

    d(xn)4. (22)

  • 8/12/2019 VJOM-D-13-00067R2

    10/19

    10 Pham N. Anh et al.

    Combining (16) with (22), we obtain

    tn x2 xn x2 tn yn2

    mn

    vn(1 mn )

    2d(xn)4.

    On the other hand, we have IB is a nonexpansive mapping. Indeed, since B is a -strongly monotone mapping and 0

  • 8/12/2019 VJOM-D-13-00067R2

    11/19

    A Fixed Point Scheme for Nonexpansive Mappings 11

    we have

    nf(xn,y) +

    1

    2y xn2 nf(x

    n,yn) +1

    2yn xn2 y C.

    Withy=xn Cand f(xn,xn) =0, we have

    0 nf(xn,yn) +

    1

    2yn xn2. (23)

    By hypothesis, f(xn, )is convex and subdifferentiable on C, i.e.,

    f(xn,y) f(xn,xn) un,y xn y C,

    whereun 2f(xn,xn). Puttingy=yn, we have

    f(xn,yn) un,yn xn.

    Combining this and (23), we obtain

    nun,yn xn +

    1

    2xn yn2 0,

    and so

    nunyn xn +12

    xn yn2 0.

    Hence

    xn yn 2nun

  • 8/12/2019 VJOM-D-13-00067R2

    12/19

    12 Pham N. Anh et al.

    Thus, we have

    xn+1 xn =PCng(xn) + (InA)Swn PCn1g(xn1) + (In1A)Swn1

    (ng(xn) + (InA)Swn) n1g(xn1) + (In1A)Swn1

    =ng(xn) ng(x

    n1) +ng(xn1) n1g(x

    n1) + (InA)Swn

    (InA)Swn1 + (InA)Sw

    n1 (In1A)Swn1

    nxn xn1 + |n n1|g(x

    n1)

    +InASwn

    Swn1

    + |n n1|ASwn1

    nx

    n xn1 + (1 n)wn wn1

    +|n n1|(g(xn1) + ASwn1)

    nxn xn1 + (1 n)x

    n xn1

    +|n n1|(g(xn1) + ASwn1)

    = (1 n( ))xn xn1 + |n n1|M, (25)

    whereM= supn0{g(xn1)+ASwn1}. Applying Lemma 6 for n:=n(),bn:=

    |nn1|n

    and using the assumptions (B1 B3), we have limn xn xn1= 0. Hence

    limn xn+1 xn =0.

    Step 8.We claim that limn xn tn =0.

    Indeed, for eachx , we have

    xn+1 x2 =PCng(xn) + (InA)Swnx2

    ng(xn) + (InA)Sw

    n x2

    =n(g(xn) Ax) +I(Swn x) nA(Sw

    n x)2

    =n(g(xn) Ax) + (InA)(Sw

    n x)2

    (ng(xn) Ax + (1 n)Sw

    n x)2

    = 2n g(xn) Ax2 + (1 n)

    2Swn x2

    +2n(1 n)g(xn) AxSwn x

    = 2n g(xn) Ax2 + (1 n)

    2Swn x2

    +2n(1 n)g(xn) AxSwn x

    2

    n (g

    (xn

    ) g

    (x

    ) + g

    (x

    ) Ax

    )

    2

    + (1

    n

    )

    2

    Swn

    x

    2

    +2n(1 n)g(xn) AxSwn x

    = 2n g(xn) g(x)2 + 22n g(x

    n) g(x)g(x) Ax

    +2n g(x) Ax2 + (1 n)

    2Swn x2

    +2n(1 n)g(xn) AxSwn x

    = 2n2g(xn) g(x)2 + 22n g(x

    n) g(x)g(x) Ax

    +2n g(x) Ax2 + (1 n)

    2Swn Sx2

    +2n(1 n)g(xn) AxSwn x

    2n22xn x2 + (1 n)

    2wn x2

    +2n g(x) Ax2 + 22n g(x

    n) g(x)g(x) Ax

    +2n(1 n)g(xn) AxSwn x

  • 8/12/2019 VJOM-D-13-00067R2

    13/19

    A Fixed Point Scheme for Nonexpansive Mappings 13

    2n2xn x2 + (1 n)

    2tn x2

    +2n g(x) Ax2 + 22n g(x

    n) g(x)g(x) Ax

    +2n(1 n)g(xn) AxSwn x

    2n2xn x2

    +(1 n)2

    xn x2 tn yn2

    mnvn(1 mn )

    2d(xn)4

    +2n g(x

    ) Ax2 + 22n g(xn) g(x)g(x) Ax

    +2n(1 n)g(xn) AxSwn x

    (2n 2 + (1 n)

    2)xn x2 (1 n)2tn yn2 +n

    xn x2 (1 n)2tn yn2 +n,

    where

    n =2n g(x

    ) Ax2 + 22n g(xn) g(x)g(x) Ax

    +2n(1 n)g(xn) AxSwn x (1 n)

    2 mn

    vn(1 mn )

    2d(xn)4.

    It then follows that

    (1 n)2tn yn2 xn x2 xn+1 x +n.

    Hence limn tn yn =0.

    On the other hand xn yn2 = mn vn,d(xn)2

    vn2 , which implies that limn xn yn =0.Since

    xn tn xn yn + tn yn,

    we obtain that limn xn tn =0.

    Step 9.We claim that limn

    xn Sxn =0.

    Indeed, by Steps 7, 8, we have

    xn+1 x2 2n g(xn) Ax2 + (1 n)

    2wn x2

    +2n(1 n)g(xn) Axwn x

    2n g(xn) Ax2 + (1 n)

    2(IB)tn (IB)x2

    +2n(1 n)g(xn) Axwn x

    2n g(xn) Ax2 + (1 n)

    2tn x2 +( 2)Btn Bx2+2n(1 n)g(x

    n) Axwn x

    2n g(xn) Ax2 + (1 n)

    2tn x2

    +(1 n)2( 2)Btn Bx2

    +2n(1 n)g(xn) Axwn x

    2n g(xn) Ax2 + (1 n)

    2tn x2

    +(1 n)2a(b 2)Btn Bx2

    +2n(1 n)g(xn) Axwn x

    2n g(xn) Ax2 + (1 n)

    2xn x2

    +(1 n)2a(b 2)Btn Bx2

    +2n(1 n)g(xn) Axwn x.

  • 8/12/2019 VJOM-D-13-00067R2

    14/19

    14 Pham N. Anh et al.

    Hence

    (1 n)2a(b 2)Btn Bx2 2n g(x

    n) Ax2 + (1 n)2xn x2

    xn+1 x2 + 2n(1 n)g(xn)

    Axwn x.

    Using(B1)and this inequality, we have limn Btn Bx =0.

    On the other hand, by Lemma 1 and Step 7, we have

    wn x2 =PC(tn Btn) PC(x Bx)2

    (tn Btn) (x Bx),wn x

    = 1

    2((tn Btn) (x Bx)2 + wn x2

    (tn Btn) (x Bx) (wn x)2)

    1

    2(tn x2 + wn x2 (tn wn) (Btn Bx)2)

    1

    2

    tn x2 + wn x2 tn wn2

    +

    1

    2(2tn wn,Btn Bx

    2Btn Bx2).

    Hence

    wn x2 tn x2 tn wn2 + 2tn wn,Btn Bx 2

    Btn Bx2.

    From this inequality and Step 8, we have

    xn+1 x2 2n g(xn) Ax2 + (1 n)

    2wn x2

    +2n(1 n)g(xn) Axwn x

    2n g(xn) Ax2 + (1 n)

    2(tn x2 tn wn2

    +2tn wn,Btn Bx 2

    Btn Bx2)

    +2n(1 n)g(xn) Axwn x

    2n g(xn) Ax2 + (1 n)

    2xn x2 (1 n)2(tn wn2

    +2(1 n)2tn wn,Btn Bx 2

    (1 n)2Btn Bx2)+2n(1 n)g(x

    n) Axwn x,

    and hence

    (1 n)2tn wn2 2n g(x

    n) Ax2 + (1 n)2xn x2 xn+1 x2

    +2(1 n)2tn wn,Btn Bx

    2(1 n)

    2Btn Bx2

    +2n(1 n)g(xn) Axwn x.

    Sincen 0 and Btn Bx 0, we obtain

    limn

    tn wn =0.

  • 8/12/2019 VJOM-D-13-00067R2

    15/19

    A Fixed Point Scheme for Nonexpansive Mappings 15

    From (8), we have

    xn+1 Swn= PC(ng(xn) + (InA)Sw

    n) Swn

    ng(xn) + (InA)Sw

    n Swn

    = ng(xn) ASwn 0, asn .

    Since

    tn Stn tn xn + xn xn+1 + xn+1 Swn + Swn Stn

    tn xn + xn xn+1 + xn+1 Swn + wn tn,

    we obtain that tn Stn 0 as n . Moreover, we get

    wn Swn wn tn + tn Stn + Stn Swn

    2wn tn + tn Stn 0 as n .

    Since

    xn wn xn tn + tn wn,

    it implies that

    limn

    xn wn =0.

    Since

    xn Sxn xn xn+1 + xn+1 Swn + Swn Sxn

    xn xn+1 + xn+1 Swn + wn xn,

    we obtain that

    limn

    xn Sxn =0.

    Step 10.We claim that

    limsupn

    (g A)q,Swn q 0. (26)

    Indeed, we choose a subsequence {wnk} of{wn} such that

    limsupn

    (g A)q,Swn q =limsupk

    (g A)q,Swnk q.

    Since {wn} is bounded, there exists a subsequence {wnki } of{wnk} which converges weaklytox. Next we show that x .

    We prove that x Fix(S). We may assume without loss of generality that wnk x.Since wn Swn 0, we obtainSwnk p. Since xn Sxn 0, xn wn 0 and byLemma 4, we havex Fix(S).

    We show thatx Sol(f,C). From Steps 7 and 10, we have

    tnk x, xnk x, ynk x.

    Sinceyn is the unique solution of the convex minimization problem

    yn =argmin

    nf(x

    n,y) +1

    2y xn2 : y C

    ,

  • 8/12/2019 VJOM-D-13-00067R2

    16/19

    16 Pham N. Anh et al.

    we have

    0 2

    nf(x

    n,y) +1

    2y xn2

    (yn) +NC(y

    n).

    It follows that

    0= nz +yn xn +zn,

    wherez 2f(xn,yn)and zn NC(yn). By the definition of the normal cone NC(yn), we getthat

    yn xn,y yn nz,yn y y C. (27)

    On the other hand, since f(xn

    , )is subdifferentiable onCand z 2f(xn

    ,yn

    ), we have

    f(xn,y) f(xn,yn) z,y yn y C. (28)

    Combining (27) with (27), we have

    n(f(xn,y) f(xn,yn)) yn xn,yn y y C.

    Hence

    nk(f(xnk,y) f(xnk,ynk)) ynk xnk,ynk y y C.

    Thus, using {n} [c,d], 0

  • 8/12/2019 VJOM-D-13-00067R2

    17/19

    A Fixed Point Scheme for Nonexpansive Mappings 17

    Step 11.We claim that xn q.Indeed, it follows from Step 2 that

    xn+1 q2 = PC(ng(xn) + (InA)Sw

    n) PC(q)2

    ng(xn) + (InA)Sw

    n q2

    2n g(xn) Aq2 + InA

    2Swn q2

    +2n(InA)(Swn q),g(xn) Aq

    2n g(xn) Aq2 + (1 n)

    2Swn q2

    +2nSwn

    q,g(xn

    ) Aq 22

    n A(Swn

    q),g(xn

    ) Aq (1 n)

    2wn q2 +2n g(xn) Aq2

    +2nSwn q,g(xn) Aq 22n A(Sw

    n q),g(xn) Aq

    (1 n)2tn q2 +2n g(x

    n) Aq2

    +2nSwn q,g(xn) g(q) + 2nSw

    n q,g(q) Aq

    22n A(Swn q),g(xn) Aq

    (1 n)2xn q2 +2n g(x

    n) Aq2 + 2nwn qxn q

    +2nSwn q,g(q) Aq 22n A(Sw

    n q),g(xn) Aq

    (1 n)2xn q2 +2n g(x

    n) Aq2 + 2nxn q2

    +2nSwn q,g(q) Aq 22n A(Sw

    n q),g(xn) Aq

    ((1 n)2 + 2n)x

    n q2 +2n g(xn) Aq2

    +2nSwn q,g(q) Aq + 22n A(Swn q)g(xn) Aq

    (1 2n())xn q2 +n(ng(x

    n) Aq2 +n2xn q2

    +2nA(Swn q)g(xn) Aq + 2Swn q,g(q) Aq).

    Otherwise {xn}, {g(xn)}, and {Swn} are bounded, we can take a constant M>0 such that

    M supn0

    {ng(xn) Aq2 +n

    2xn q2 + 2nA(Swn q)g(xn) Aq}.

    This implies that

    xn+1 q2 (1 2n())xn q2 +nn, (29)

    wheren= 2Swn q,g(q) Aq +n M. From (26), it follows that limsupnn 0.

    Applying Lemma 6 for (29), we obtain that xn

    qasn . This completes the proof.

    From Algorithm 1, we can easily deduce the algorithm for solving the equilibrium prob-

    lem EP(f,C)as follows.

    Algorithm 2 Choose positive sequences {n} (0,1), [a,b],0

  • 8/12/2019 VJOM-D-13-00067R2

    18/19

    18 Pham N. Anh et al.

    Step 2. Find the smallest positive integer number mnsuch that

    f(xn mn d(xn),yn) d(xn)2.

    Compute

    tn =PCHn (xn),

    where zn =xn mn d(xn),vn 2f(zk,zk) and Hn= {x H : v

    n,x zn 0}.Increase n by1 and go back to Step1, and go to Step3.

    Step 3. Compute

    xn+1 =PC(ng(xn) + (InA)(tn)).

    Increase n by1 and go back to Step1.

    As in Theorem 1, we give a convergent result of Algorithm 2 as the following.

    Theorem 2 LetH be a real Hilbert space and C be a closed convex subset of H. Let

    f :C C R be a bifunction satisfying (A1)(A4) and A be a strongly positive linearbounded operator ofH into itself with coefficient> 0such thatA = 1. Let g : C C be

    a contraction with coefficient (0,1). Assume that0

  • 8/12/2019 VJOM-D-13-00067R2

    19/19

    A Fixed Point Scheme for Nonexpansive Mappings 19

    11. Goebel, K., Kirk, W.A.: Topics in Metric Fixed Point Theory. Cambridge Studies in Advanced Mathe-

    matics 28. Cambridge University Press, Cambridge (1990)

    12. Marino, G., Xu, H.-K.: A general iterative method for nonexpansive mappings in Hilbert spaces. J. Math.

    Anal. Appl.318, 4352 (2006)

    13. Mastroeni, G.: Gap functions for equilibrium problems. J. Glob. Optim.27, 411426 (2003)14. Peng, J.-W., Yao, J.-C.: Some new extragradient-like methods for generalized equilibrium problems,

    fixed point problems and variational inequality problems. Optim. Methods Softw.25, 677698 (2010)

    15. Rockafellar, R.T.: On the maximality of sums of nonlinear monotone operators. Trans. Am. Math. Soc.149, 7588 (1970)

    16. Takahashi, S., Takahashi, W.: Viscosity approximation methods for equilibrium problems and fixed pointproblems in Hilbert spaces. J. Math. Anal. Appl.331, 506515 (2007)

    17. Wangkeeree, R., Preechasilp, P.: A new iterative scheme for solving the equilibrium problems, variational

    inequality problems, and fixed point problems in Hilbert spaces. J. Appl. Math., 154968 (2012), 21 pp.

    doi:10.1155/2012/154968

    18. Wang, S., Cho, Y.J., Qin, X.: A new iterative method for solving equilibrium problems and fixed

    point problems for an infinite family of nonexpansive mappings. Fixed Point Theory Appl. (2010). doi:

    10.1155/2010/165098

    19. Wang, S., Guo, B.: New iterative scheme with nonexpansive mappings for equilibrium problems and

    variational inequality problems in Hilbert spaces. J. Comput. Appl. Math. 233, 26202630 (2010)

    20. Xu, H.K.: An iterative approach to quadratic optimization. J. Optim. Theory Appl.116, 659678 (2003)