The convergence of mann iteration for an asymptotic hemicontractive map

Abstract

We prove that the Mann iteration convergence ot a fixed point of an asymptotic hemicontractive map.

Authors

Stefan M. Soltuz
Tiberiu Popoviciu Institute of Numerical analysis

Keywords

Mann iteration; fixed pont; asymptotic hemicontractive map

Paper coordinates

Ș.M. Șoltuz, The convergence of mann iteration for an asymptotic hemicontractive map, Buletinul ştiinţific al Universitatii Baia Mare, Seria B, Fascicola matematică-informatică, 18 (2002) no. 1, pp. 115-118 (Dedicated to Costica Mustata on his 60th anniversary).

PDF

About this paper

Journal

Bull. Stiint. Univ. Baia Mare, Matematica-Informatica

Publisher Name

Baia Mare University

DOI
Print ISSN

1222-1201 

Online ISSN

google scholar link

??

Paper (preprint) in HTML form

2002-Soltuz-Bull-Baia-Mare-The-convergence-of-mann-iteration
Dedeculed to Costica MUSTATA on his 60 th annwersary

THE CONVERGENCE OF MANN ITERATION FOR AN ASYMPTOTIC HEMICONTRACTIVE MAP

Ştefan M. ŞOLTUZ

Abstract. We prove that the Mam iteration converges to a fixed point of an asymptotic hemicontractive map.
MSC: 47H10
Keywords: Mann iteration, fixed point, asymptotic hemicontractive map

1. Introduction.

Let X X XXX be a real Hilbert space, let B X B X B sub XB \subset XBX be a nonempty, convex set. Lot T : B B T : B B T:B rarr BT: B \rightarrow BT:BB be a map. Let N = { 1 , 2 , } N = { 1 , 2 , } N={1,2,dots}\mathbb{N}=\{1,2, \ldots\}N={1,2,}. Let x 1 B x 1 B x_(1)in Bx_{1} \in Bx1B, be an arbitrary fixed point. We consider the iteration
(1) x n + 1 = ( 1 α n ) x n + α n T n x n , n N . (1) x n + 1 = 1 α n x n + α n T n x n , n N . {:(1)x_(n+1)=(1-alpha_(n))x_(n)+alpha_(n)T^(n)x_(n)","AA n in N.:}\begin{equation*} x_{n+1}=\left(1-\alpha_{n}\right) x_{n}+\alpha_{n} T^{n} x_{n}, \forall n \in N . \tag{1} \end{equation*}(1)xn+1=(1αn)xn+αnTnxn,nN.
The sequence ( α n ) n 1 α n n 1 (alpha_(n))_(n >= 1)\left(\alpha_{n}\right)_{n \geq 1}(αn)n1 satisfies:
( α n ) n 1 < ( 0 , 1 ) , n = 1 α n = , lim n α n = 0 α n n 1 < ( 0 , 1 ) , n = 1 α n = , lim n α n = 0 (alpha_(n))_(n >= 1) < (0,1),sum_(n=1)^(oo)alpha_(n)=oo,lim_(n rarr oo)alpha_(n)=0\left(\alpha_{n}\right)_{n \geq 1}<(0,1), \sum_{n=1}^{\infty} \alpha_{n}=\infty, \lim _{n \rightarrow \infty} \alpha_{n}=0(αn)n1<(0,1),n=1αn=,limnαn=0
A prototype for ( κ n ) n 1 κ n n 1 (kappa_(n))_(n >= 1)\left(\kappa_{n}\right)_{n \geq 1}(κn)n1 is ( 1 / n ) n 1 ( 1 / n ) n 1 (1//n)_(n >= 1)(1 / n)_{n \geqslant 1}(1/n)n1. Iteration (1) is known as Mam iteration, sox [ 8 ] [ 8 ] [8][8][8]. We consider the following iteration, known as Ishikawa iteration, soe [6]:
(Ish) x n + 1 = ( 1 α n ) x n + α n T n y n + 1 y n = ( 1 β n ) x n + β n T n x n + 1 n N , (Ish) x n + 1 = 1 α n x n + α n T n y n + 1 y n = 1 β n x n + β n T n x n + 1 n N , {:[(Ish)x_(n+1)=(1-alpha_(n))x_(n)+alpha_(n)T^(n)y_(n+1)],[y_(n)=(1-beta_(n))x_(n)+beta_(n)T^(n)x_(n+1)quad n inN","]:}\begin{gather*} x_{n+1}=\left(1-\alpha_{n}\right) x_{n}+\alpha_{n} T^{n} y_{n+1} \tag{Ish}\\ y_{n}=\left(1-\beta_{n}\right) x_{n}+\beta_{n} T^{n} x_{n+1} \quad n \in \mathbb{N}, \end{gather*}(Ish)xn+1=(1αn)xn+αnTnyn+1yn=(1βn)xn+βnTnxn+1nN,
where ( α n ) n , ( β n ) n ( 0 , 1 ) α n n , β n n ( 0 , 1 ) (alpha_(n))_(n),(beta_(n))_(n)sub(0,1)\left(\alpha_{n}\right)_{n},\left(\beta_{n}\right)_{n} \subset(0,1)(αn)n,(βn)n(0,1). Choosing β n = 0 , n N β n = 0 , n N beta_(n)=0,AA n inN\beta_{n}=0, \forall n \in \mathbb{N}βn=0,nN, from ( T s h T s h TshT s hTsh ) we get (1).
Let us denote by F ( T ) F ( T ) F(T)F(T)F(T) the set of fixed points of the operator T T TTT. We need the following . definition, see for example [10]:
Definition 1. The map T : B B T : B B T:B rarr BT: B \rightarrow BT:BB is culled asymptotically bemicumtractme with sequence ( k n ) n k n n (k_(n))_(n)\left(k_{n}\right)_{n}(kn)n, if and only if lim n , k n , 1 , F ( T ) lim n , k n , 1 , F ( T ) lim_(n rarr oo),k_(n),-1,F(T)!=O/\lim _{n \rightarrow \infty}, k_{n},-1, F(T) \neq \emptysetlimn,kn,1,F(T) such that
(2) T n x x 2 k n x x 2 + x T n x 2 , x B , x F ( T ) , n N . (2) T n x x 2 k n x x 2 + x T n x 2 , x B , x F ( T ) , n N . {:(2)||T^(n)x-x^(**)||^(2) <= k_(n)||x-x^(**)||^(2)+||x-T^(n)x||^(2)","AA x in B","x^(**)in F(T)","AA n inN.:}\begin{equation*} \left\|T^{n} x-x^{*}\right\|^{2} \leq k_{n}\left\|x-x^{*}\right\|^{2}+\left\|x-T^{n} x\right\|^{2}, \forall x \in B, x^{*} \in F(T), \forall n \in \mathbb{N} . \tag{2} \end{equation*}(2)Tnxx2knxx2+xTnx2,xB,xF(T),nN.
In this note, we will consider
k n < 1 , n N . k n < 1 , n N . k_(n) < 1,AA n inN.k_{n}<1, \forall n \in \mathbb{N} .kn<1,nN.
In context of Hilbert spaces, the convergence of Ishikawa iteration (I ah) to the fixed point of T T TTT, when we deal with s asymptotically hemicontractive map (with k n > 1 , V n k n > 1 , V n k_(n) > 1,Vn ink_{n}>1, \mathrm{~V} n \inkn>1, Vn N), could be found in [10]. A convergence result in normed spaces for (Ish) (with ( α n ) n , ( β n ) n ( 0 , 1 ) α n n , β n n ( 0 , 1 ) (alpha_(n))_(n),(beta_(n))_(n)sub(0,1)\left(\alpha_{n}\right)_{n},\left(\beta_{n}\right)_{n} \subset(0,1)(αn)n,(βn)n(0,1) being not convergent to zero), could be found in [0]. Anyway, in [5], we deal with an asymptotic hemicontractive like map, which is not the same as in Definition 2 with condition (2). Let us remark that:
(i) In [ 10 ] [ 10 ] [10][10][10] and ( 5 ) ( 5 ) (5)(5)(5), the sequentes ( α n ) n ( β n ) n ( 0 , 1 ) α n n β n n ( 0 , 1 ) (alpha_(n))_(n)(beta_(n))_(n)sub(0,1)\left(\alpha_{n}\right)_{n}\left(\beta_{n}\right)_{n} \subset(0,1)(αn)n(βn)n(0,1) are not convergent to zero, because of the cxistonce of an c > 0 c > 0 c > 0c>0c>0 such that c α n β n , n N c α n β n , n N c <= alpha_(n) <= beta_(n,)AA n inNc \leq \alpha_{n} \leq \beta_{n,} \forall n \in \mathbb{N}cαnβn,nN. From the convergence of (Ish) (which is proved in [10] for Hilbert spaces and T T TTT asymptotic hemicontractive), with that sequences ( α n ) n , ( β n ) n α n n , β n n (alpha_(n))_(n),(beta_(n))_(n)\left(\alpha_{n}\right)_{n},\left(\beta_{n}\right)_{n}(αn)n,(βn)n, we can not dednces convergence of Mann iteration. It is impossible to bave β e = 0 , n N β e = 0 , n N beta_(e)=0,AA n inN\beta_{e}=0, \forall n \in \mathbb{N}βe=0,nN.
(ii) According to [5], convergence results of (Ish) for asymptotic hemicontractive maps, exist only in [10], (and of course in [5]). So far as we know, no other papers are dealing with asymptotic hemicontractive maps and Mann iteration (1).
This two reasons lead us to remark a lack of a convergenoe result which deal with Mann iteration (1) for an bsymptotic hemicontractive map. In particular for a map asymptotically hemicontractive with sequence ( k n ) n , k n < 1 k n n , k n < 1 (k_(n))_(n),k_(n) < 1\left(k_{n}\right)_{n}, k_{n}<1(kn)n,kn<1, ''n N N in N\in NN. Our aim is to prove a result which deal with the convergence of Mamn-iteration (1) for an asymptotic hemicontractive map as in Definition 1.
The following lemma can be found in [11] as Lemma 4. Also, it can be found in [12] as Lemma 1.2, with an other proof. In | 1 | | 1 | |1||1||1| can be found as Lemma 2.
Lemma 1 [1]. , [11]. [12] Let ( α n ) n 1 α n n 1 (alpha_(n))_(n >= 1)\left(\alpha_{n}\right)_{n \geq 1}(αn)n1 be a nonnegotive sequence which verzfies
a n + 1 ( 1 λ n ) a n + σ n , a n + 1 1 λ n a n + σ n , a_(n+1) <= (1-lambda_(n))a_(n)+sigma_(n),a_{n+1} \leq\left(1-\lambda_{n}\right) a_{n}+\sigma_{n},an+1(1λn)an+σn,
where: ( λ n ) n > 1 ( 0.1 ) , π = 1 λ n = cos λ n n > 1 ( 0.1 ) , π = 1 λ n = cos (lambda_(n))_(n > 1)sub(0.1),sum_(pi=1)^(oo)lambda_(n)=cos\left(\lambda_{n}\right)_{n>1} \subset(0.1), \sum_{\pi=1}^{\infty} \lambda_{n}=\cos(λn)n>1(0.1),π=1λn=cos and σ n = o ( λ n ) σ n = o λ n sigma_(n)=o(lambda_(n))\sigma_{n}=o\left(\lambda_{n}\right)σn=o(λn). Then lim n , a n 0 n , a n 0 n rarr oo,a_(n)!=0n \rightarrow \infty, a_{n} \neq 0n,an0.
The following result is from 61 :
Lemma 2 [6]. Let X X XXX be a real Hilbert space, the following relation is frue for all x , y X x , y X x,y in Xx, y \in Xx,yX, and for all λ ( 0 , 1 ) λ ( 0 , 1 ) lambda in(0,1)\lambda \in(0,1)λ(0,1) :
(3) ( 1 λ ) x + λ y 2 = ( 1 λ ) x 2 + λ y 2 λ ( 1 λ ) x y 2 (3) ( 1 λ ) x + λ y 2 = ( 1 λ ) x 2 + λ y 2 λ ( 1 λ ) x y 2 {:(3)||(1-lambda)x+lambda y||^(2)=(1-lambda)||x||^(2)+lambda||y||^(2)-lambda(1-lambda)||x-y||^(2):}\begin{equation*} \|(1-\lambda) x+\lambda y\|^{2}=(1-\lambda)\|x\|^{2}+\lambda\|y\|^{2}-\lambda(1-\lambda)\|x-y\|^{2} \tag{3} \end{equation*}(3)(1λ)x+λy2=(1λ)x2+λy2λ(1λ)xy2

2. Main result

We are able now to give the following result:
Theorem 1. Let X X XXX he a real Hilbert syme and let B X B X B sub XB \subset XBX be a nonempty convex bounded set, and T : B B T : B B T:B rarr BT: B \rightarrow BT:BB be an asymptotic homicontractive map with ( α n ) n 1 k n < 1 α n n 1 k n < 1 (alpha_(n))_(n >= 1)k_(n) < 1\left(\alpha_{n}\right)_{n \geq 1} k_{n}<1(αn)n1kn<1. Let { x } = F ( T ) x = F ( T ) {x^(**)}=F(T)\left\{x^{*}\right\}=F(T){x}=F(T). Ir ( α n ) n α n n (alpha_(n))_(n)\left(\alpha_{n}\right)_{n}(αn)n given by (1) verifies α n = o ( ( 1 k n ) ) α n = o 1 k n alpha_(n)=o((1-k_(n)))\alpha_{n}=o\left(\left(1-k_{n}\right)\right)αn=o((1kn)), then the Mann iteration ( x n ) n 1 x n n 1 (x_(n))_(n >= 1)\left(x_{n}\right)_{n \geq 1}(xn)n1, given by ( 1 ) ( 1 ) (1)(1)(1), is comprgent to x x x^(**)x^{*}x, for all x 1 B x 1 B x_(1)in Bx_{1} \in Bx1B.
Proof. The sequence ( x n x ) n x n x n (||x_(n)-x^(**)||)_(n)\left(\left\|x_{n}-x^{*}\right\|\right)_{n}(xnx)n is woll-defined, bexause x x x^(**)x^{*}x is unique. Using (3) and (2) we have, w'v M M inM\in \mathrm{M}M,
x n + 1 x 2 ( 1 α n ) ( x n x ) + α n ( T n x n T n x ) | 2 = ( 1 α n ) x n x n 2 + α n | T n x n x n 2 α n ( 1 α n ) T n x n x n 2 ( 1 α n ) x n x 2 + α n ( x n x n x 2 + T n x n x n 2 ) = ( 1 α n ( 1 α n ) T n x n x n 2 α n ( 1 α n ) x n n 2 + α n k n x n x n 2 + α n T n x n x n 2 = ( 1 α n ( 1 x n ) ) x n x n 2 + α n 2 T n x n x n 2 ( 1 α n ( 1 k n ) ) x n x n 2 + α n 2 M x n + 1 x 2 1 α n x n x + α n T n x n T n x 2 = 1 α n x n x n 2 + α n T n x n x n 2 α n 1 α n T n x n x n 2 1 α n x n x 2 + α n x n x n x 2 + T n x n x n 2 = 1 α n 1 α n T n x n x n 2 α n 1 α n x n n 2 + α n k n x n x n 2 + α n T n x n x n 2 = 1 α n 1 x n x n x n 2 + α n 2 T n x n x n 2 1 α n 1 k n x n x n 2 + α n 2 M {:[||x_(n+1)-x^(**)||^(2)(1-alpha_(n))(x_(n)-x^(**))+alpha_(n)(T^(n)x_(n)-T^(n)x^(**))|^(2)],[=(1-alpha_(n))||x_(n)-x^(n)||^(2)+alpha_(n)|T^(n)x_(n)-x^(n)||^(2)-alpha_(n)(1-alpha_(n))∣T^(n)x_(n)-x_(n)||^(2):}],[ <= (1-alpha_(n))||x_(n)-x^(**)||^(2)+alpha_(n)(x_(n)||x_(n)-x^(**)||^(2)+∣T^(n)x_(n)-x_(n)||^(2))],[=(1-alpha_(n)(1-alpha_(n))∣T^(n)x_(n)-x_(n)||^(2):}],[-alpha_(n)(1-alpha_(n))||x^(n)||_(n)^(2)+alpha_(n)k_(n)||x_(n)-x_(n)||^(2)+alpha_(n)||T^(n)x_(n)-x_(n)||^(2)],[=(1-alpha_(n)(1-x_(n)))||x_(n)-x^(n)||^(2)+alpha_(n)^(2)||T^(n)x_(n)-x_(n)||^(2)],[ <= (1-alpha_(n)(1-k_(n)))||x_(n)-x^(n)||^(2)+alpha_(n)^(2)M]:}\begin{aligned} \left\|x_{n+1}-x^{*}\right\|^{2} & \left(1-\alpha_{n}\right)\left(x_{n}-x^{*}\right)+\left.\alpha_{n}\left(T^{n} x_{n}-T^{n} x^{*}\right)\right|^{2} \\ = & \left(1-\alpha_{n}\right)\left\|x_{n}-x^{n}\right\|^{2}+\alpha_{n}\left|T^{n} x_{n}-x^{n}\left\|^{2}-\alpha_{n}\left(1-\alpha_{n}\right) \mid T^{n} x_{n}-x_{n}\right\|^{2}\right. \\ \leq & \left(1-\alpha_{n}\right)\left\|x_{n}-x^{*}\right\|^{2}+\alpha_{n}\left(x_{n}\left\|x_{n}-x^{*}\right\|^{2}+\mid T^{n} x_{n}-x_{n} \|^{2}\right) \\ = & \left(1-\alpha_{n}\left(1-\alpha_{n}\right) \mid T^{n} x_{n}-x_{n} \|^{2}\right. \\ & -\alpha_{n}\left(1-\alpha_{n}\right)\left\|x^{n}\right\|_{n}^{2}+\alpha_{n} k_{n}\left\|x_{n}-x_{n}\right\|^{2}+\alpha_{n}\left\|T^{n} x_{n}-x_{n}\right\|^{2} \\ = & \left(1-\alpha_{n}\left(1-x_{n}\right)\right)\left\|x_{n}-x^{n}\right\|^{2}+\alpha_{n}^{2}\left\|T^{n} x_{n}-x_{n}\right\|^{2} \\ \leq & \left(1-\alpha_{n}\left(1-k_{n}\right)\right)\left\|x_{n}-x^{n}\right\|^{2}+\alpha_{n}^{2} M \end{aligned}xn+1x2(1αn)(xnx)+αn(TnxnTnx)|2=(1αn)xnxn2+αn|Tnxnxn2αn(1αn)Tnxnxn2(1αn)xnx2+αn(xnxnx2+Tnxnxn2)=(1αn(1αn)Tnxnxn2αn(1αn)xnn2+αnknxnxn2+αnTnxnxn2=(1αn(1xn))xnxn2+αn2Tnxnxn2(1αn(1kn))xnxn2+αn2M
The last inequality is true, since the sequence ( T n x n x n 2 ) n 1 T n x n x n 2 n 1 (||T^(n)x_(n)-x_(n)||^(2))_(n >= 1)\left(\left\|T^{n} x_{n}-x_{n}\right\|^{2}\right)_{n \geq 1}(Tnxnxn2)n1 is bounded, because B B BBB is bounded. There exists M > 0 M > 0 M > 0M>0M>0 such that T n x n x n 2 < M T n x n x n 2 < M ||T^(n)x_(n)-x_(n)||^(2) < M\left\|T^{n} x_{n}-x_{n}\right\|^{2}<MTnxnxn2<M, for all n N n N n inNn \in \mathbb{N}nN. We denote by a n := x n x , λ n = α n ( 1 k n ) , σ n = α n 2 M , n N a n := x n x , λ n = α n 1 k n , σ n = α n 2 M , n N a_(n):=||x_(n)-x^(**)||,lambda_(n)=alpha_(n)(1-k_(n)),sigma_(n)=alpha_(n)^(2)M,AA n inNa_{n}:=\left\|x_{n}-x^{*}\right\|, \lambda_{n}=\alpha_{n}\left(1-k_{n}\right), \sigma_{n}=\alpha_{n}^{2} M, \forall n \in \mathbb{N}an:=xnx,λn=αn(1kn),σn=αn2M,nN. Because α n = o ( ( 1 k n ) ) α n = o 1 k n alpha_(n)=o((1-k_(n)))\alpha_{n}= o\left(\left(1-k_{n}\right)\right)αn=o((1kn)), we have lim n κ n 1 k n M = 0 lim n κ n 1 k n M = 0 lim_(n rarr oo)(kappa_(n))/(1-k_(n))M=0\lim _{n \rightarrow \infty} \frac{\kappa_{n}}{1-k_{n}} M=0limnκn1knM=0. All the conditions from Lemma 2 are fulfilled. Thus lim n a n = 0 lim n a n = 0 lim_(n rarr oo)a_(n)=0\lim _{n \rightarrow \infty} a_{n}=0limnan=0. The proof is complete.
In Theorem 4 we don't need any Lipschitx condition for T T TTT as in [10]

REFERENCES.

[1] S.S.Chung, Y.J. Cho, B.S. Lee, J.S. Jung, S. M. Kang, Iterative Approximutions of Fixed Poinls and Solutuons for Strongly Acoretive and Strongity Psendo-contractive Mappirigs in Banach Spucres, J. Math. Anal. Appl. 224 (1998), 149-165.
12 S.S. Chang, J.Y. Park, Y.J. Cho, Iterulive Approminations of Fixed Points Jow Asymetotically Noncrpansive Mappings in Barach Spaces, Bull. Korean Math. Snc. 37 (2000), 109-119.
[3] S.S. Chang: Sonc Reaults for Asymptotwally Paeudo-Coniractive Mappings and Asymptotically Nonerpansive Mappings, Proc. Atwer. Math. Suc, 129 (2000), 845-853.
[4] S. S. Chang, Y. J. Cho, Haiyun Zhou, Demi-closed Principle and Weuk Convergence Problems for Assymptoticaliy Nonerpansiec Mappinga, J. Korean Math. Soc. 38 (2001):6, 1245-1260.
[5] C. E. Chidume, Convergence Theorems for Asymptotically Pseudocontractive Moppings, Nonlinear Anal. 49 (2002), 1-11.
6 S. Ishikawa, Fized Points by a New Theratow Method, Proc. Amer. Math. Soc. 44 (1974), 147-150.
(7) Gang Eun KIM, Tae Hwa KIM, Mame And Ishifown Thernivens with Errora for Non-Lipschitaion mappings in Banach Spaces, Computers and Muth. Appl 42 (2001), 1565-1570.
[8] W. R. Mann, Mean Value an Itenation, Proc. Amer. Math. Sod. 4 (1953), 506-510.
[9] M.O. Osilike, S.C. Anigbosor, Weak and Strong Convergence Theorems for Fized Points Of Asymptoticolly Noneryansvive Mappings Muth. Computer Modeling 32 (2000), 1181-1191.
10'. L. Qihou, Contergence Theomens of the Sequence of Therales for asymptoticully Demicontrantive and Hemicontractive Mappings, Nonlinear Anal. 26 (1996):11, 18351842.
[11] X. Weng, Fired Pomit Iteration for Locol Structly Pseudocontracticue Mapping, Proc. Amer. Math. Soc. 113 (1991), 727-731.
[12] H. Zhou, J. Yuting, Approximation of Fzzed Points of Strongly Pseadocontractive Maps without Lipschilz Assumption, Proc. Amer. Math. Soc. 125 (1997), 1705-1709.
Receivext: 21.04.2002
Kurt Schumaher str., ur. 48, ap. 38, 67663 Kaiserslautern,
Germany.
E-mail: ssoltuzdyahoo.com
ssoliuxthotmail.com
soltuz itwm.fhg.de
2002

Related Posts