The convergence of modified Mann-Ishikawa iterations when applied to an asymptotically pseudocontractive map

Abstract

We prove that under minimal conditions the modified Mann and Ishikawa iterations converge when dealing with an asymptotically pseudocontractive map. We give an affirmative answer to the open question from C.E. Chidume and H. Zegeye, Approximate fixed point sequences and convergence theorems for asymptotically pseudocontractive mappings, J. Math. Anal. Appl., 278 (2003), 354–366.

Authors

Stefan M. Soltuz
Tiberiu Popoviciu Institute of Numerical Analysis, Romanian Academy

Keywords

Asymptotically hemicontractive map; Modified Mann; Modified Ishikawa iteration

Paper coordinates

Stefan M. Soltuz, The convergence of modified Mann-Ishikawa iterations when applied to an asymptotically pseudocontractive map, Austral. J. Math Anal. Appl., Volume 4, Issue 2, Article 16, pp. 1-8, 2007.

PDF

About this paper

Journal

The Australian Journal of Mathematical
Analysis and Applications

Publisher Name
Print ISSN

1449-5910

Online ISSN

google scholar link

[1] S.S. CHANG, Some results for asymptotically pseudo-contractive mappings and asymptotically nonexpansive mappings, Proc. Amer. Math. Soc., 129 (2000), 845–853.
[2] S.S. CHANG, J.Y. PARK and Y.J. CHO, Iterative approximations of fixed points for asymptotically nonexpansive mappings in Banach spaces, Bull. Korean Math. Soc., 37 (2000), 109–119.
[3] C.E. CHIDUME, Convergence theorems for asymptotically pseudocontractive mappings, Nonlinear Analysis, 49 (2002), 1–11.
[4] C.E. CHIDUME and H. ZEGEYE, Approximate fixed point sequences and convergence theorems for asymptotically pseudocontractive mappings, J. Math. Anal. Appl., 278 (2003), 354–366.
[5] D.I. IGBOKWE, Iterative construction of fixed points of asymptotically pseudocontractive maps, Panamer. Math. J., 13 (2003), 83–97.
[6] S. ISHIKAWA, Fixed points by a new iteration method, Proc. Amer. Math. Soc., 44 (1974), 147-150.
[7] T. KATO, Nonlinear semigroup and evolution equations, J. Math. Soc. Japan, 19(1967), 508–520.
[8] W.R. MANN, Mean value in iteration, Proc. Amer. Math. Soc., 4 (1953), 506–510.
[9] M.O. OSILIKE, Iterative approximation of fiexd points of asymptotically demicontractive mappings, Indian J. Pure Appl. Math., 29 (1998), 1291–1300.
[10] M.O. OSILIKE and D.I. IGBBOKWE, Convergence theorems for asymptotically pseudocontractive maps, Bull. Korean Math. Soc., 39 (2002), 389–399.
[11] B.E. RHOADES and ¸STEFAN M. ¸SOLTUZ, The equivalence between the convergences of Ishikawa and Mann iterations for asymptotically pseudocontractive map, J. Math. Anal. Appl., 283 (2003), 681–688.
[12] B.K. SHARMA and D.R. SAHU, Existence and approximation results for asymptotically pseudocontractive mappings, Indian J. Pure Appl. Math., 31 (2000), 185–196.
[13] J. SCHU, Iterative construction of fixed points of asymptotically nonexpansive mappings, J. Math. Anal. Appl., 158 (1991), 407–413.
[14] H. ZHOU and J. YUTING, Approximation of fixed points of strongly pseudocontractive maps without Lipschitz assumption, Proc. Amer. Math. Soc., 125 (1997), 1705–1709.

Paper (preprint) in HTML form

THE CONVERGENCE OF MODIFIED MANN-ISHIKAWA ITERATIONS WHEN APPLIED TO AN ASYMPTOTICALLY PSEUDOCONTRACTIVE MAP

Departamento de Matematicas, Universidad de Los Andes, Carrera 1, No. 18A-10, Bogota, Colombia, and "T. Popoviciu" Institute of Numerical Analysis, Cluj-Napoca, Romania, smsoltuz@gmail.com
URL:http://www.uniandes.edu.co
Abstract

We prove that under minimal conditions the modified Mann and Ishikawa iterations converge when dealing with an asymptotically pseudocontractive map. We give an affirmative answer to the open question from C.E. Chidume and H. Zegeye, Approximate fixed point sequences and convergence theorems for asymptotically pseudocontractive mappings, J. Math. Anal. Appl., 278 (2003), 354-366.

Received 26 January, 2007; accepted 5 September, 2007; published 30 November, 2007. Communicated by: M. Mariani

1. Introduction

Let XX be an arbitrary real Banach space and J:X2XJ:X\rightarrow 2^{X^{*}} the normalized duality mapping given by

Jx:={fX:x,f=xf,f=x},xX.Jx:=\left\{f\in X^{*}:\langle x,f\rangle=\|x\|\|f\|,\|f\|=\|x\|\right\},\forall x\in X. (1.1)

In [13] the following class of maps was introduced:
Definition 1.1. Let XX be a normed space and BB a subset of XX. A map TT is said to be asymptotically pseudocontractive if there exists a sequence {Kn},Kn[1,),n,limnKn=\left\{K_{n}\right\},K_{n}\in[1,\infty),\forall n\in\mathbb{N},\lim_{n\rightarrow\infty}K_{n}= 1 , and there exists j(xy)J(xy)j(x-y)\in J(x-y) such that

TnxTny,j(xy)Knxy2,x,yB,n.\left\langle T^{n}x-T^{n}y,j(x-y)\right\rangle\leq K_{n}\|x-y\|^{2},\forall x,y\in B,\forall n\in\mathbb{N}. (1.2)

If there exists xx^{*} such that Tx=xTx^{*}=x^{*}, by setting y:=xy:=x^{*} in (1.2) we get

Tnxx,j(xx)Knxx2,x,yB,n;\left\langle T^{n}x-x^{*},j\left(x-x^{*}\right)\right\rangle\leq K_{n}\left\|x-x^{*}\right\|^{2},\forall x,y\in B,\forall n\in\mathbb{N}; (1.3)

such a map is called asymptotically hemicontractive.

The modified Mann iteration, (see [8]), is defined by

un+1=(1αn)un+αnTnun.u_{n+1}=\left(1-\alpha_{n}\right)u_{n}+\alpha_{n}T^{n}u_{n}. (1.4)

The modified Ishikawa iteration is defined, (see [6]), by

xn+1\displaystyle x_{n+1} =(1αn)xn+αnTnyn,\displaystyle=\left(1-\alpha_{n}\right)x_{n}+\alpha_{n}T^{n}y_{n}, (1.5)
yn\displaystyle y_{n} =(1βn)xn+βnTnxn.\displaystyle=\left(1-\beta_{n}\right)x_{n}+\beta_{n}T^{n}x_{n}.

The sequences {αn}(0,1),{βn}[0,1)\left\{\alpha_{n}\right\}\subset(0,1),\left\{\beta_{n}\right\}\subset[0,1) satisfy

limnαn=limnβn=0,n=1αn=+.\lim_{n\rightarrow\infty}\alpha_{n}=\lim_{n\rightarrow\infty}\beta_{n}=0,\sum_{n=1}^{\infty}\alpha_{n}=+\infty. (1.6)

We shall give here the most general result concerning the convergence of Mann and Ishikawa iterations dealing with a uniformly Lipschitzian and asymptotically pseudocontractive map. Our result generalizes the main results from [3], [4], [10] and [12]. We also give an affirmative answer to the open question from [4] if the Mann or Ishikawa iteration converges when applied to an asymptotically pseudocontractive (respectively an asymptotically hemicontractive map), in more general spaces than Hilbert spaces.

2. Preliminaries

We recall the following auxiliary results.
Lemma 2.1. [7] Let XX be a Banach space and x,yXx,y\in X. Then

xx+ry\|x\|\leq\|x+ry\| (2.1)

for all r>0r>0 if and only if there exists j(x)J(x)j(x)\in J(x) such that y,j(x)0\langle y,j(x)\rangle\geq 0.

Lemma 2.2 [11] Let BB be a nonempty subset of a Banach space XX and let T:BBT:B\rightarrow B be a map. Then the following conditions are equivalent:
(i) TT is an asymptotically pseudocontractive map,
(ii) for kn[1,),nk_{n}\in[1,\infty),\forall n\in\mathbb{N}, we have

xyxy+r[(knITn)x(knITn)y],x,yB,r>0.\|x-y\|\leq\left\|x-y+r\left[\left(k_{n}I-T^{n}\right)x-\left(k_{n}I-T^{n}\right)y\right]\right\|,\forall x,y\in B,\forall r>0. (2.2)

Definition 2.1. Let XX be a normed space and BB a subset of XX, then the map T:BBT:B\rightarrow B is a uniformly Lipschitzian map if for some L1L\geq 1, we have TnxTnyLxy,x,yB,n\left\|T^{n}x-T^{n}y\right\|\leq L\|x-y\|,\forall x,y\in B,\forall n\in\mathbb{N}.

Lemma 2.3. [14] Let {Ψn}\left\{\Psi_{n}\right\} be a nonnegative sequence satisfying

Ψn+1(1λn)Ψn+σn\Psi_{n+1}\leq\left(1-\lambda_{n}\right)\Psi_{n}+\sigma_{n} (2.3)

where λn(0,1),n=1λn=+\lambda_{n}\in(0,1),\sum_{n=1}^{\infty}\lambda_{n}=+\infty and σn=o(λn)\sigma_{n}=o\left(\lambda_{n}\right). Then limnΨn=0\lim_{n\rightarrow\infty}\Psi_{n}=0.

3. Main Result

Theorem 3.1. Let BB be a closed convex subset of an arbitrary Banach space XX and (un)n\left(u_{n}\right)_{n} defined by (1.4) with (αn)n\left(\alpha_{n}\right)_{n} and (βn)n\left(\beta_{n}\right)_{n} satisfying (1.6). Let TT be an asymptotically pseudocontractive (or asymptotically hemicontractive) and uniformly Lipschitzian map with L1L\geq 1 self-map of BB. If u0Bu_{0}\in B, then the modified Mann iteration (1.4) strongly converges to the nearest xx^{*} fixed point of TT.

Proof. From (1.4) we obtain

un\displaystyle u_{n} =un+1+αnunαnTnun\displaystyle=u_{n+1}+\alpha_{n}u_{n}-\alpha_{n}T^{n}u_{n} (3.1)
=(1+αn2)un+1+αn(αnknITn)un+1+\displaystyle=\left(1+\alpha_{n}^{2}\right)u_{n+1}+\alpha_{n}\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}+
(1+kn)αn2un+1+αnun+αn(Tnun+1Tnun)\displaystyle-\left(1+k_{n}\right)\alpha_{n}^{2}u_{n+1}+\alpha_{n}u_{n}+\alpha_{n}\left(T^{n}u_{n+1}-T^{n}u_{n}\right)
=(1+αn2)un+1+αn(αnknITn)un+1+\displaystyle=\left(1+\alpha_{n}^{2}\right)u_{n+1}+\alpha_{n}\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}+
(1+kn)αn2[un+αn(Tnunun)]+αnun+αn(Tnun+1Tnun)\displaystyle-\left(1+k_{n}\right)\alpha_{n}^{2}\left[u_{n}+\alpha_{n}\left(T^{n}u_{n}-u_{n}\right)\right]+\alpha_{n}u_{n}+\alpha_{n}\left(T^{n}u_{n+1}-T^{n}u_{n}\right)
=(1+αn2)un+1+αn(αnknITn)un+1+(1+kn)αn3(unTnun)+\displaystyle=\left(1+\alpha_{n}^{2}\right)u_{n+1}+\alpha_{n}\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}+\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+
+[1(1+kn)αn]αnun+αn(Tnun+1Tnun)\displaystyle+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}u_{n}+\alpha_{n}\left(T^{n}u_{n+1}-T^{n}u_{n}\right)

By using Tnx=xT^{n}x^{*}=x^{*} we observe that

x=(1+αn2)x+αn(αnknITn)x+[1(1+kn)αn]αnxx^{*}=\left(1+\alpha_{n}^{2}\right)x^{*}+\alpha_{n}\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}x^{*} (3.2)

From (3.1) and (3.2) we get

xun\displaystyle x^{*}-u_{n} (3.3)
=(1+αn2)(xun+1)+\displaystyle=\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+
+αn((αnknITn)x(αnknITn)un+1)+\displaystyle+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)+
+[1(1+kn)αn]αn(xun)+(1+kn)αn3(Tnunun)+\displaystyle+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\left(x^{*}-u_{n}\right)+\left(1+k_{n}\right)\alpha_{n}^{3}\left(T^{n}u_{n}-u_{n}\right)+
+αn(TnunTnun+1)\displaystyle+\alpha_{n}\left(T^{n}u_{n}-T^{n}u_{n+1}\right)

The norm of the sum of the first two terms on the right-hand side of 3.3) is equal to

(1+αn2)(xun+1)+αn1+αn2((αnknITn)x(αnknITn)un+1).\left(1+\alpha_{n}^{2}\right)\left\|\left(x^{*}-u_{n+1}\right)+\frac{\alpha_{n}}{1+\alpha_{n}^{2}}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)\right\|. (3.4)

Using (2.1) with

x\displaystyle x :=xun+1,\displaystyle:=x^{*}-u_{n+1}, (3.5)
y\displaystyle y :=(αnknITn)x(αnknITn)un+1,\displaystyle:=\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1},
r\displaystyle r :=αn1+αn2,\displaystyle:=\frac{\alpha_{n}}{1+\alpha_{n}^{2}},

we obtain

(1+αn2)(xun+1)+αn((αnknITn)x(αnknITn)un+1)\displaystyle\left\|\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)\right\| (3.6)
(1+αn2)xun+1.\displaystyle\geq\left(1+\alpha_{n}^{2}\right)\left\|x^{*}-u_{n+1}\right\|.

From (3.3) and (3.6) it follows that

xun\displaystyle\left\|x^{*}-u_{n}\right\| (3.7)
?(1+αn2)(xun+1)+αn((αnknITn)x(αnknITn)un+1)+\displaystyle\stackrel{{\scriptstyle?}}\left\|\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)\right\|+
+[1(1+kn)αn]αnxun(1+kn)αn3Tnunun+\displaystyle+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\left\|x^{*}-u_{n}\right\|-\left(1+k_{n}\right)\alpha_{n}^{3}\left\|T^{n}u_{n}-u_{n}\right\|+
αnTnunTnun+1\displaystyle-\alpha_{n}\left\|T^{n}u_{n}-T^{n}u_{n+1}\right\|
(1+αn2)xun+1+[1(1+kn)αn]αnxun+\displaystyle\geq\left(1+\alpha_{n}^{2}\right)\left\|x^{*}-u_{n+1}\right\|+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\left\|x^{*}-u_{n}\right\|+
(1+kn)αn3TnununαnTnunTnun+1\displaystyle-\left(1+k_{n}\right)\alpha_{n}^{3}\left\|T^{n}u_{n}-u_{n}\right\|-\alpha_{n}\left\|T^{n}u_{n}-T^{n}u_{n+1}\right\|

We shall prove later the first inequality from 3.7). Supposing that 3.7) holds, we obtain

(1+αn2)xun+1\displaystyle\left(1+\alpha_{n}^{2}\right)\left\|x^{*}-u_{n+1}\right\| (3.8)
{1[1(1+kn)αn]αn}xun+(1+kn)αn3Tnunun+\displaystyle\leq\left\{1-\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\right\}\left\|x^{*}-u_{n}\right\|+\left(1+k_{n}\right)\alpha_{n}^{3}\left\|T^{n}u_{n}-u_{n}\right\|+
+αnTnunTnun+1\displaystyle+\alpha_{n}\left\|T^{n}u_{n}-T^{n}u_{n+1}\right\|

Also, we know that

unTnun\displaystyle\left\|u_{n}-T^{n}u_{n}\right\| TnunTnx+xun\displaystyle\leq\left\|T^{n}u_{n}-T^{n}x^{*}\right\|+\left\|x^{*}-u_{n}\right\| (3.9)
Lxun+xun\displaystyle\leq L\left\|x^{*}-u_{n}\right\|+\left\|x^{*}-u_{n}\right\|
=(L+1)xun\displaystyle=(L+1)\left\|x^{*}-u_{n}\right\|

Using (1.4), 3.9 and the fact that TT is a uniformly Lipschitzian map, we obtain

Tnun+1Tnun\displaystyle\left\|T^{n}u_{n+1}-T^{n}u_{n}\right\| Lun+1un\displaystyle\leq L\left\|u_{n+1}-u_{n}\right\| (3.10)
=αnLunTnun\displaystyle=\alpha_{n}L\left\|u_{n}-T^{n}u_{n}\right\|
αnL(L+1)xun\displaystyle\leq\alpha_{n}L(L+1)\left\|x^{*}-u_{n}\right\|

From (3.8), (3.9) and (3.10), by using (1+αn2)11,n\left(1+\alpha_{n}^{2}\right)^{-1}\leq 1,\forall n\in\mathbb{N}, we get

xun+1\displaystyle\left\|x^{*}-u_{n+1}\right\| (3.11)
{1[1(1+kn)αn]αn}xun+\displaystyle\leq\left\{1-\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\right\}\left\|x^{*}-u_{n}\right\|+
+(1+kn)αn3(L+1)xun+αn2L(L+1)xun\displaystyle+\left(1+k_{n}\right)\alpha_{n}^{3}(L+1)\left\|x^{*}-u_{n}\right\|+\alpha_{n}^{2}L(L+1)\left\|x^{*}-u_{n}\right\|

The condition limnαn=0\lim_{n\rightarrow\infty}\alpha_{n}=0 implies the existence of n0n_{0}\in\mathbb{N}, such that

αn2118(1+L),nn0\alpha_{n}^{2}\leq\frac{1}{18(1+L)},\forall n\geq n_{0} (3.12)

Condition (3.12) assures the following inequalities, nn0\forall n\geq n_{0},

αn\displaystyle\alpha_{n} 13((1+kn)+L(1+L))\displaystyle\leq\frac{1}{3\left(\left(1+k_{n}\right)+L(1+L)\right)} (3.13)
αn2\displaystyle\alpha_{n}^{2} 13(1+L)(1+kn)\displaystyle\leq\frac{1}{3(1+L)\left(1+k_{n}\right)}
αn\displaystyle\alpha_{n} 13\displaystyle\leq\frac{1}{3}

Using (3.11) and (3.13) we observe that

{1[1(1+kn)αn]αn}+(1+kn)αn3(L+1)+αn2L(L+1)\displaystyle\left\{1-\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\right\}+\left(1+k_{n}\right)\alpha_{n}^{3}(L+1)+\alpha_{n}^{2}L(L+1) (3.14)
=1αn+αn[((1+kn)+L(1+L))αn+(1+kn)(L+1)αn2]\displaystyle=1-\alpha_{n}+\alpha_{n}\left[\left(\left(1+k_{n}\right)+L(1+L)\right)\alpha_{n}+\left(1+k_{n}\right)(L+1)\alpha_{n}^{2}\right]
1αn+αn(13+13)=113αn\displaystyle\leq 1-\alpha_{n}+\alpha_{n}\left(\frac{1}{3}+\frac{1}{3}\right)=1-\frac{1}{3}\alpha_{n}

Relations (3.11), (3.14), lead us to

xun+1(113αn)xun,nn0\left\|x^{*}-u_{n+1}\right\|\leq\left(1-\frac{1}{3}\alpha_{n}\right)\left\|x^{*}-u_{n}\right\|,\forall n\geq n_{0} (3.15)

Setting in (2.3) from lemma 2.3

Ψn\displaystyle\Psi_{n} :=xun,nn0\displaystyle:=\left\|x^{*}-u_{n}\right\|,\forall n\geq n_{0} (3.16)
λn\displaystyle\lambda_{n} :=13αn,nn0\displaystyle:=\frac{1}{3}\alpha_{n},\forall n\geq n_{0}
σn\displaystyle\sigma_{n} :=0,n\displaystyle:=0,\forall n\in\mathbb{N}

we get

limnΨn=limnxun=0\lim_{n\rightarrow\infty}\Psi_{n}=\lim_{n\rightarrow\infty}\left\|x^{*}-u_{n}\right\|=0 (3.17)

We prove now the first inequality from 3.7 . Set in 3.7

a\displaystyle a =(1+αn2)(xun+1)\displaystyle=\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right) (3.18)
a\displaystyle a^{\prime} =(1+αn2)(xun+1)+\displaystyle=\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+
+αn((αnknITn)x(αnknITn)un+1)\displaystyle+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)
b\displaystyle b =[1(1+kn)αn]αn(xun)\displaystyle=\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\left(x^{*}-u_{n}\right)
c\displaystyle c =(1+kn)αn3(Tnunun)\displaystyle=\left(1+k_{n}\right)\alpha_{n}^{3}\left(T^{n}u_{n}-u_{n}\right)
d\displaystyle d =αn(Tnun+1Tnun)\displaystyle=\alpha_{n}\left(T^{n}u_{n+1}-T^{n}u_{n}\right)

to obtain

a+b+c+d\displaystyle\left\|a^{\prime}+b+c+d\right\| (3.19)
=xun=(1+αn2)(xun+1)+\displaystyle=\left\|x^{*}-u_{n}\right\|=\|\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+
+αn((αnknITn)x(αnknITn)un+1)+\displaystyle+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)+
+[1(1+kn)αn]αn(xun)+(1+kn)αn3(Tnunun)+\displaystyle+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\left(x^{*}-u_{n}\right)+\left(1+k_{n}\right)\alpha_{n}^{3}\left(T^{n}u_{n}-u_{n}\right)+
+αn(TnunTnun+1)\displaystyle+\alpha_{n}\left(T^{n}u_{n}-T^{n}u_{n+1}\right)\|
?(1+αn2)(xun+1)+αn((αnknITn)x(αnknITn)un+1)+\displaystyle\stackrel{{\scriptstyle?}}\left\|\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)\right\|+
+[1(1+kn)αn]αnxun(1+kn)αn3Tnunun+\displaystyle+\left[1-\left(1+k_{n}\right)\alpha_{n}\right]\alpha_{n}\left\|x^{*}-u_{n}\right\|-\left(1+k_{n}\right)\alpha_{n}^{3}\left\|T^{n}u_{n}-u_{n}\right\|+
αnTnunTnun+1\displaystyle-\alpha_{n}\left\|T^{n}u_{n}-T^{n}u_{n+1}\right\|
=a+bcd\displaystyle=\left\|a^{\prime}\right\|+\|b\|-\|c\|-\|d\|

We shall now prove (3.19) using the following relations

a+b+c+d+c+d\displaystyle\left\|a^{\prime}+b+c+d\right\|+\|c\|+\|d\| (3.20)
a+b+c+d+c+d\displaystyle\geq\left\|a^{\prime}+b+c+d\right\|+\|c+d\|
?a+b\displaystyle\stackrel{{\scriptstyle?}}\left\|a^{\prime}\right\|+\|b\|
a+b.\displaystyle\geq\|a\|+\|b\|.

The last inequality from (3.20) is given by (2.1), that is aa\left\|a^{\prime}\right\|\geq\|a\|. We further prove that

a+b+c+d+c+d?a+b.\left\|a^{\prime}+b+c+d\right\|+\|c+d\|\stackrel{{\scriptstyle?}}\left\|a^{\prime}\right\|+\|b\|. (3.21)

By using

xun+1=xun+αn(unTnun),x^{*}-u_{n+1}=x^{*}-u_{n}+\alpha_{n}\left(u_{n}-T^{n}u_{n}\right), (3.22)

we obtain

a\displaystyle\left\|a^{\prime}\right\| =(1+αn2)(xun+1)+\displaystyle=\|\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+ (3.23)
+αn((αnknITn)x(αnknITn)un+1)\displaystyle+\alpha_{n}\left(\left(\alpha_{n}k_{n}I-T^{n}\right)x^{*}-\left(\alpha_{n}k_{n}I-T^{n}\right)u_{n+1}\right)\|
=(1+αn2)(xun+1)+knαn2(xun+1)\displaystyle=\|\left(1+\alpha_{n}^{2}\right)\left(x^{*}-u_{n+1}\right)+k_{n}\alpha_{n}^{2}\left(x^{*}-u_{n+1}\right)
αn(TnxTnun+1)\displaystyle-\alpha_{n}\left(T^{n}x^{*}-T^{n}u_{n+1}\right)\|
=(1+kn)αn2(αnunαnTnun)+\displaystyle=\|\left(1+k_{n}\right)\alpha_{n}^{2}\left(\alpha_{n}u_{n}-\alpha_{n}T^{n}u_{n}\right)+
+(1+kn)αn2(xun)+(xun+1)+\displaystyle+\left(1+k_{n}\right)\alpha_{n}^{2}\left(x^{*}-u_{n}\right)+\left(x^{*}-u_{n+1}\right)+
αn(TnxTnun+1)\displaystyle-\alpha_{n}\left(T^{n}x^{*}-T^{n}u_{n+1}\right)\|
=(1+kn)αn3(unTnun)+\displaystyle=\|\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+
αn(TnxTnun+1)+(1+kn)αn2(xun)+\displaystyle-\alpha_{n}\left(T^{n}x^{*}-T^{n}u_{n+1}\right)+\left(1+k_{n}\right)\alpha_{n}^{2}\left(x^{*}-u_{n}\right)+
+(xun)+αn(unTnun)\displaystyle+\left(x^{*}-u_{n}\right)+\alpha_{n}\left(u_{n}-T^{n}u_{n}\right)\|
=(1+kn)αn3(unTnun)+\displaystyle=\|\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+
αn(TnxTnun+1Tnx+Tnun)+(1+kn)αn2(xun)+\displaystyle-\alpha_{n}\left(T^{n}x^{*}-T^{n}u_{n+1}-T^{n}x^{*}+T^{n}u_{n}\right)+\left(1+k_{n}\right)\alpha_{n}^{2}\left(x^{*}-u_{n}\right)+
+(xun)+αn(unTnun)+αn(Tnx+Tnun)\displaystyle+\left(x^{*}-u_{n}\right)+\alpha_{n}\left(u_{n}-T^{n}u_{n}\right)+\alpha_{n}\left(-T^{n}x^{*}+T^{n}u_{n}\right)\|
=(1+kn)αn3(unTnun)+\displaystyle=\|\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+
αn(Tnun+1+Tnun)+(1+kn)αn2(xun)+\displaystyle-\alpha_{n}\left(-T^{n}u_{n+1}+T^{n}u_{n}\right)+\left(1+k_{n}\right)\alpha_{n}^{2}\left(x^{*}-u_{n}\right)+
+(xun)+αn(unx).\displaystyle+\left(x^{*}-u_{n}\right)+\alpha_{n}\left(u_{n}-x^{*}\right)\|.

The last equality is true because Tnx=xT^{n}x^{*}=x^{*}. Finally, we have

a\displaystyle\left\|a^{\prime}\right\| =(1+kn)αn3(unTnun)+\displaystyle=\|\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+ (3.24)
αn(TnunTnun+1)+\displaystyle-\alpha_{n}\left(T^{n}u_{n}-T^{n}u_{n+1}\right)+
+(1αn+(1+kn)αn2)(xun)\displaystyle+\left(1-\alpha_{n}+\left(1+k_{n}\right)\alpha_{n}^{2}\right)\left(x^{*}-u_{n}\right)\|
((1+kn)αn3(unTnun)+αn(TnunTnun+1))+\displaystyle\leq\left\|-\left(\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+\alpha_{n}\left(T^{n}u_{n}-T^{n}u_{n+1}\right)\right)\right\|+
+(1αn+(1+kn)αn2)xun\displaystyle+\left(1-\alpha_{n}+\left(1+k_{n}\right)\alpha_{n}^{2}\right)\left\|x^{*}-u_{n}\right\|
=xun+\displaystyle=\left\|x^{*}-u_{n}\right\|+
+((1+kn)αn3(unTnun)+αn(TnunTnun+1))+\displaystyle+\left\|-\left(\left(1+k_{n}\right)\alpha_{n}^{3}\left(u_{n}-T^{n}u_{n}\right)+\alpha_{n}\left(T^{n}u_{n}-T^{n}u_{n+1}\right)\right)\right\|+
(1(1+kn)αn2)xun\displaystyle-\left(1-\left(1+k_{n}\right)\alpha_{n}^{2}\right)\left\|x^{*}-u_{n}\right\|
=a+b+c+d+c+db.\displaystyle=\left\|a^{\prime}+b+c+d\right\|+\|c+d\|-\|b\|.

The last equality is true because we already know that

xun=a+b+c+d.\left\|x^{*}-u_{n}\right\|=\left\|a^{\prime}+b+c+d\right\|. (3.25)

Remark 3.1. If limnαn0\lim_{n\rightarrow\infty}\alpha_{n}\neq 0, then our Theorem 3.1 holds supposing condition (3.12) is satisfied.

The modified Ishikawa iteration also converges, being equivalent to the modified Mann iteration.

Theorem 3.2. [11] Let BB be a closed convex subset of an arbitrary Banach space X,(xn)nX,\left(x_{n}\right)_{n} and (un)n\left(u_{n}\right)_{n} defined by (1.5) and (1.4) with (αn)n\left(\alpha_{n}\right)_{n} and (βn)n\left(\beta_{n}\right)_{n} satisfying (1.6). Let TT be an asymptotically pseudocontractive and uniformly Lipschitzian with L1L\geq 1 self-map of BB. Let xx^{*} be a fixed point of TT. If u0=x0Bu_{0}=x_{0}\in B, then the following two assertions are equivalent:
(i) the modified Mann iteration (1.4) strongly converges to xx^{*},
(ii) the modified Ishikawa iteration (1.5) strongly converges to xx^{*}.

Remark 3.2. Each fixed point has its own basin of attraction. The map TT has no unique fixed point. The starting point is crucial for the convergence of Mann or Ishikawa iteration. For example, take T=IT=I, the identity map on BB, with kn=1,nk_{n}=1,\forall n\in\mathbb{N}. Each point of BB becomes a fixed point and the starting point is directly a fixed point.

Theorem 3.1 generalizes the Theorem from [3] because in [3] the set BB is bounded, the space XX is uniformly convex and (αn)n\left(\alpha_{n}\right)_{n} and (βn)n\left(\beta_{n}\right)_{n} satisfy some additional conditions. We also generalize Theorem 1 from [12], because the space is smooth and the following conditions are required: (kn1)<+,αn2<+\sum\left(k_{n}-1\right)<+\infty,\sum\alpha_{n}^{2}<+\infty and βn<+\sum\beta_{n}<+\infty. Our Theorem 3.1 generalizes the main results from [2] and [10] because the map TT satisfies the following restrictive condition:

Tnxn+1x,j(xn+1x)knxn+1x2ϕ(xn+1x),\left\langle T^{n}x_{n+1}-x^{*},j\left(x_{n+1}-x^{*}\right)\right\rangle\leq k_{n}\left\|x_{n+1}-x^{*}\right\|^{2}-\phi\left(\left\|x_{n+1}-x^{*}\right\|\right), (3.26)

where (xn)n\left(x_{n}\right)_{n} is the modified Mann (respectively modified Ishikawa) iterations, xx^{*} is a fixed point and ϕ:[0,)[0,)\phi:[0,\infty)\rightarrow[0,\infty) is a strictly increasing function with ϕ(0)=0\phi(0)=0.

In [1] and [5] the convergence of (1.4) and (1.5) is shown, dealing with an asymptotically pseudocontractive map without being uniformly Lipschitzian. However, in [1] and [5] the assumptions are more restrictive than those from our Theorem 3.1; the Banach space is uniformly smooth, the set BB is bounded, respectively T(B)T(B) is bounded and the map TT satisfies condition (3.26).

References

[1] S.S. CHANG, Some results for asymptotically pseudo-contractive mappings and asymptotically nonexpansive mappings, Proc. Amer. Math. Soc., 129 (2000), 845-853.
[2] S.S. CHANG, J.Y. PARK and Y.J. CHO, Iterative approximations of fixed points for asymptotically nonexpansive mappings in Banach spaces, Bull. Korean Math. Soc., 37 (2000), 109-119.
[3] C.E. CHIDUME, Convergence theorems for asymptotically pseudocontractive mappings, Nonlinear Analysis, 49 (2002), 1-11.
[4] C.E. CHIDUME and H. ZEGEYE, Approximate fixed point sequences and convergence theorems for asymptotically pseudocontractive mappings, J. Math. Anal. Appl., 278 (2003), 354-366.
[5] D.I. IGBOKWE, Iterative construction of fixed points of asymptotically pseudocontractive maps, Panamer. Math. J., 13 (2003), 83-97.
[6] S. ISHIKAWA, Fixed points by a new iteration method, Proc. Amer. Math. Soc., 44 (1974), 147150.
[7] T. KATO, Nonlinear semigroup and evolution equations, J. Math. Soc. Japan, 19(1967), 508-520.
[8] W.R. MANN, Mean value in iteration, Proc. Amer. Math. Soc., 4 (1953), 506-510.
[9] M.O. OSILIKE, Iterative approximation of fiexd points of asymptotically demicontractive mappings, Indian J. Pure Appl. Math., 29 (1998), 1291-1300.
[10] M.O. OSILIKE and D.I. IGBBOKWE, Convergence theorems for asymptotically pseudocontractive maps, Bull. Korean Math. Soc., 39 (2002), 389-399.
[11] B.E. RHOADES and ŞTEFAN M. ŞOLTUZ, The equivalence between the convergences of Ishikawa and Mann iterations for asymptotically pseudocontractive map, J. Math. Anal. Appl., 283 (2003), 681-688.
[12] B.K. SHARMA and D.R. SAHU, Existence and approximation results for asymptotically pseudocontractive mappings, Indian J. Pure Appl. Math., 31 (2000), 185-196.
[13] J. SCHU, Iterative construction of fixed points of asymptotically nonexpansive mappings, J. Math. Anal. Appl., 158 (1991), 407-413.
[14] H. ZHOU and J. YUTING, Approximation of fixed points of strongly pseudocontractive maps without Lipschitz assumption, Proc. Amer. Math. Soc., 125 (1997), 1705-1709.

2007

Related Posts