The equivalence of Mann and Ishikawa iterations dealing with uniformly pseudocontractive maps without bounded range

(original), paper

Abstract

We prove that Mann and Ishikawa iterations are equivalent models dealing with \(\psi\)-uniformly pseudocontractive or d-weakly contractive maps without bounded range

Authors

Stefan M. Soltuz
Tiberiu Popoviciu Institute of Numerical Analysis

B.E. Rhoades
Department of Mathematics, Indiana University, Bloomington, IN 47405-7106, U.S.A

Keywords

\(\psi\)-uniformly pseudocontractive maps; d-weakly contractivemap;~ Mannand Ishikawa iterations

Paper coordinates

B.E. Rhoades, Ş.M. Şoltuz, The equivalence of Mann and Ishikawa iterations dealing with uniformly pseudocontractive maps without bounded range, Tamkang J. Math. 37 (3) (2006).

PDF

About this paper

Journal

Tamkang Journal of Mathematics

Publisher Name

Tamkang University Tamsui,  Taiwan, R.O.C.

Print ISSN

0049-2930

Online ISSN

2073-9826

google scholar link

[1] Ya. I. Alber and S. Reich, An iterative method for solving a class of nonlinear operatorequations in Banach spaces,Panamer. Math. J.4(1994), 39-54.
[2] Ya. I. Alber and S. Guerre-Delabriere,Principle of weakly contractive maps in Hilbertspaces, Operator Theory98(1997), 7-22.
[3] Ya. I. Alber and S. Guerre-Delabriere,On the projection methods for fixed point problems,Analysis21(2001), 17-39.
[4] M. S. Berger, Nonlinearity and Functional Analysis, Academic Press, New York, 1977, page65.
[5] C. E. Chidume, H. Zegeye and S. J. Aneke,Approximation of fixed points of weakly con-tractive nonself maps in Banach spaces, J. Math. Anal. Appl.270(2002), 189-199.
[6] C. E. Chidume and H. Zegeye,Approximation methods for nonlinear operator equations,Proc. Amer. Math. Soc.131(2003), 2467-2478.
[7] K. Deimling, Nonlinear Functional Analysis, Springer-Verlag, 1985, page 115.
[8] S. Ishikawa,Fixed points by a new iteration method, Proc. Amer. Math. Soc.44(1974),147-150.
[9] W. R. Mann,Mean value in iteration, Proc. Amer. Math. Soc.4(1953), 506-510.
[10] C. Morales and J. S. Jung,Convergence of paths for pseudocontractive mappings in Banachspaces,Proc. Amer. Math. Soc.128(2000), 3411-3419.
[11] B. E. Rhoades and S. M. S ̧oltuz,On the equivalence of Mann and Ishikawa iterationmethods,Int. J. Math. Math. Sci.2003(7), 451-459.
[12] B. E. Rhoades and S. M. S ̧oltuz,The equivalence of Mann and Ishikawa iteration forψuniformly pseudocontractive orψuniformly accretive maps,Int. J. Math. Math. Sci.46(2004), 2443-2452.

Paper (preprint) in HTML form

THE EQUIVALENCE OF MANN AND ISHIKAWA ITERATIONS DEALING WITH 𝝍\boldsymbol{\psi}-UNIFORMLY PSEUDOCONTRACTIVE MAPS WITHOUT BOUNDED RANGE

B. E. RHOADES AND ŞTEFAN M. ŞOLTUZ
Abstract

We prove that Mann and Ishikawa iterations are equivalent models dealing with ψ\psi-uniformly pseudocontractive or d-weakly contractive maps without bounded range.

1. Introduction

In this paper XX denotes a real Banach space with XX^{*} strictly convex, T:XXT:X\rightarrow X a map and let x0,u0Xx_{0},u_{0}\in X. We consider the following iteration known as Mann iteration, ([9])

un+1=(1αn)un+αnTun.u_{n+1}=\left(1-\alpha_{n}\right)u_{n}+\alpha_{n}Tu_{n}. (1.1)

The sequence {αn}(0,1)\left\{\alpha_{n}\right\}\subset(0,1) satisfies limnαn=0\lim_{n\rightarrow\infty}\alpha_{n}=0, and n=1αn=\sum_{n=1}^{\infty}\alpha_{n}=\infty. We consider the following iteration known as Ishikawa iteration, ([8])

xn+1\displaystyle x_{n+1} =(1αn)xn+αnTyn,\displaystyle=\left(1-\alpha_{n}\right)x_{n}+\alpha_{n}Ty_{n}, (1.2)
yn\displaystyle y_{n} =(1βn)xn+βnTxn.\displaystyle=\left(1-\beta_{n}\right)x_{n}+\beta_{n}Tx_{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 (1.3)

The duality normalized map J:X2XJ:X\rightarrow 2^{X^{*}} is given by

J(x)={fX:f,x=x2,x=f}J(x)=\left\{f\in X^{*}:\langle f,x\rangle=\|x\|^{2},\|x\|=\|f\|\right\} (1.4)

We have

f,yfy,yX\langle f,y\rangle\leq\|f\|\|y\|,\forall y\in X (1.5)

The following Remark is Proposition 12.3 from [7].

00footnotetext: Received April 14, 2005; revised March 9, 2006.
2000 Mathematics Subject Classification. 47H10.
Key words and phrases. ψ\psi-uniformly pseudocontractive maps, d-weakly contractive map, Mann and Ishikawa iterations.

Remark 1.1.([7]) If XX is a real Banach space with XX * strictly convex then J()J(\cdot) is a single map and uniformly continuous on all the bounded sets of XX.

The following result is Lemma 1 from [10].
Lemma 1.2. If XX is a real normed space, then the following relation is true

x+y2x2+2y,j(x+y),x,yX,j(x+y)J(x+y).\|x+y\|^{2}\leq\|x\|^{2}+2\langle y,j(x+y)\rangle,\forall x,y\in X,\forall j(x+y)\in J(x+y). (1.6)

The following definitions are from [3], [5] and [6].
Definition 1.3. Let XX be a normed space.
A map T:XXT:X\rightarrow X is called weakly contractive map if for all x,yXx,y\in X, there exist ψ:[0,+)[0,+)\psi:[0,+\infty)\rightarrow[0,+\infty) a continuous and strictly increasing map such that ψ\psi is positive on (0,+),ψ(0)=0(0,+\infty),\psi(0)=0, and the following inequality is satisfied

TxTyxyψ(xy)\|Tx-Ty\|\leq\|x-y\|-\psi(\|x-y\|) (1.7)

A map T:XXT:X\rightarrow X is called d-weakly contractive map if for all x,yXx,y\in X, there exist j(xy)J(xy)j(x-y)\in J(x-y) and ψ:[0,+)[0,+)\psi:[0,+\infty)\rightarrow[0,+\infty) a continuous and strictly increasing map such that ψ\psi is positive on (0,+),ψ(0)=0(0,+\infty),\psi(0)=0, and the following inequality is satisfied

|TxTy,j(xy)|xy2ψ(xy)|\langle Tx-Ty,j(x-y)\rangle|\leq\|x-y\|^{2}-\psi(\|x-y\|) (1.8)

A map T:XXT:X\rightarrow X is called ψ\psi-uniformly pseudocontractive if there exist j(xy)J(xy)j(x-y)\in J(x-y) and ψ:[0,+)[0,+)\psi:[0,+\infty)\rightarrow[0,+\infty) a strictly increasing map such that ψ\psi is positive on (0,+),ψ(0)=0(0,+\infty),\psi(0)=0 and the following inequality is satisfied

TxTy,j(xy)xy2ψ(xy),x,yB.\langle Tx-Ty,j(x-y)\rangle\leq\|x-y\|^{2}-\psi(\|x-y\|),\forall x,y\in B. (1.9)

A map C:XXC:X\rightarrow X is called ψ\psi-uniformly accretive if there exist j(xy)J(xy)j(x-y)\in J(x-y) and ψ:[0,+)[0,+)\psi:[0,+\infty)\rightarrow[0,+\infty) a strictly increasing map such that ψ\psi is positive on (0,+)(0,+\infty), ψ(0)=0\psi(0)=0 and the following inequality is satisfied

CxCy,j(xy)ψ(xy),x,yX.\langle Cx-Cy,j(x-y)\rangle\geq\psi(\|x-y\|),\forall x,y\in X. (1.10)

We denote the identity map by II.
Remark 1.4. (i) If TT is a d-weakly contractive map, then TT is a ψ\psi-uniformly pseudocontractive map.
(ii) The map TT is ψ\psi-uniformly pseudocontractive if and only if C:=(IT)C:=(I-T) is ψ\psi-uniformly accretive.

Proposition 1.5. If TT is a weakly contractive map, then TT is a ψ\psi-uniformly pseudocontractive map.

Proof. Let j(xy)J(xy)j(x-y)\in J(x-y). Using (1.5), (1.7) and (1.4) we get

TxTy,j(xy)\displaystyle\langle Tx-Ty,j(x-y)\rangle TxTyj(xy)\displaystyle\leq\|Tx-Ty\|\|j(x-y)\|
=TxTyxyxy2xyϕ(xy)\displaystyle=\|Tx-Ty\|\|x-y\|\leq\|x-y\|^{2}-\|x-y\|\phi(\|x-y\|) (1.11)

Denote ψ(a):=aϕ(a),a[0,)\psi(a):=a\cdot\phi(a),\forall a\in[0,\infty) to obtain that ψ\psi is strictly increasing and positive.
The convergence of Mann iteration for a d-weakly contractive map in Hilbert spaces, was studied in [3]. It was shown in [5] that Mann iteration (1.1) for a d-weakly contractive map without a bounded range, converges in a Banach space more general then a Hilbert space. Also, it was shown in [6] that the same iteration for a ψ\psi-uniformly pseudocontractive map without a bounded range, converges in a normed space.

If TT is a weakly contractive, then TT is a nonexpansive map. In this case the equivalence between Mann and Ishikawa iterations follows from Theorem 3 of the paper [11].

The above two motivations lead us to prove, in this note, the equivalence between Mann and Ishikawa iterations, (1.1) and (1.2), dealing with ψ\psi-uniformly pseudocontractive maps without bounded range. As a corollary we obtain the convergence of Ishikawa iteration for the above operatorial classes. Also, we give a positive answer to the following conjecture, (see [11], page 452), "If Mann iteration converges, so does Ishikawa iteration".

For a ψ\psi-uniformly pseudocontractive (respectively, ψ\psi-uniformly accretive) map, the equivalence between Mann and Ishikawa iterations was shown also in Theorem 2.1 and Corollary 3.1 from [12]. There, in [12], the set T(X)T(X) was assumed to be bounded. Removing the boundedness of the range, forces us to pay a price: both {αn}\left\{\alpha_{n}\right\} and {βn}\left\{\beta_{n}\right\} will depend on TT and xx^{*} ( see condition (2.1)).

Remark 1.6. Let XX be a normed space and T:XXT:X\rightarrow X a uniformly continuous map. Then ITI-T is a uniformly continuous map.

The following result is Proposition 2.1.2 from [4].
Proposition 1.7([4]) Let XX be a normed space and T:XXT:X\rightarrow X be a uniformly continuous map. Then TT is bounded; i.e. it maps any bounded set into a bounded set.

Remark 1.6 and Proposition 1.7 lead to the following result.
Remark 1.8. Let XX be a normed space and T:XXT:X\rightarrow X a uniformly continuous map. Then ITI-T is bounded; i.e. it maps any bounded set into a bounded set.

The following result, stated below, is Lemma 3.1 from [1]. In [1], the map ψ\psi is assumed to be continuous in order to obtain an estimate for the convergence rate of the sequence {λn}\left\{\lambda_{n}\right\}. Another proof for the Lemma 3.1 can be found in ([2], pages 12-13). The same lemma, without the continuity assumption on ψ\psi, appears in [6].

Lemma 1.9.([1]) Let {λn}\left\{\lambda_{n}\right\} and {γn}\left\{\gamma_{n}\right\} be sequences of nonnegative numbers and {αn}\left\{\alpha_{n}\right\} a sequence of positive numbers satisfying the conditions n=1αn=+\sum_{n=1}^{\infty}\alpha_{n}=+\infty and (γn/αn)0\left(\gamma_{n}/\alpha_{n}\right)\rightarrow 0
as n+n\rightarrow+\infty. Suppose that

λn+1λn2αnψ(λn)+γn,\lambda_{n+1}\leq\lambda_{n}-2\alpha_{n}\psi\left(\lambda_{n}\right)+\gamma_{n}, (1.12)

is satisfied, where ψ:[0,+)[0,+)\psi:[0,+\infty)\rightarrow[0,+\infty) is a strictly increasing map such that ψ\psi is positive on (0,+)(0,+\infty), with ψ(0)=0\psi(0)=0. Then limnλn=0\lim_{n\rightarrow\infty}\lambda_{n}=0.

2. Main Result

Let F(T)F(T) denote the fixed point set of TT.
Theorem 2.1. Let XX be a real Banach space with XX^{*} stricly convex. If T:XXT:X\rightarrow X is a ψ\psi-uniformly pseudocontractive and uniformly continuous map with xF(T)x^{*}\in F(T), x0=u0Xx_{0}=u_{0}\in X and there exists a constant d0:=d0(T,x)(0,1)d_{0}:=d_{0}\left(T,x^{*}\right)\in(0,1), which depends on TT and xx^{*}, such that {αn},{βn}\left\{\alpha_{n}\right\},\left\{\beta_{n}\right\} satisfy

αn,βnd0,n,\alpha_{n},\beta_{n}\leq d_{0},\forall n\in\mathbb{N}, (2.1)

and (1.3), then the following are equivalent:
(i) the Mann iteration (1.1) converges to the xF(T)x^{*}\in F(T),
(ii) the Ishikawa iteration (1.2) converges to the same xx^{*}.

Proof. The fixed point xx^{*} is unique. If not, then there exists at least another fixed point yF(T)y^{*}\in F(T), with xyx^{*}\neq y^{*}. Relation (1.9) leads to

TxTy,J(xy)\displaystyle\left\langle Tx^{*}-Ty^{*},J\left(x^{*}-y^{*}\right)\right\rangle xy2ψ(xy)\displaystyle\leq\left\|x^{*}-y^{*}\right\|^{2}-\psi\left(\left\|x^{*}-y^{*}\right\|\right)
xy,J(xy)\displaystyle\left\langle x^{*}-y^{*},J\left(x^{*}-y^{*}\right)\right\rangle xy2ψ(xy)\displaystyle\leq\left\|x^{*}-y^{*}\right\|^{2}-\psi\left(\left\|x^{*}-y^{*}\right\|\right)
xy2\displaystyle\left\|x^{*}-y^{*}\right\|^{2} xy2ψ(xy)\displaystyle\leq\left\|x^{*}-y^{*}\right\|^{2}-\psi\left(\left\|x^{*}-y^{*}\right\|\right)
ψ(xy)\displaystyle\psi\left(\left\|x^{*}-y^{*}\right\|\right) 0xy=0\displaystyle\leq 0\Rightarrow\left\|x^{*}-y^{*}\right\|=0 (2.2)

The implication (ii) \Rightarrow (i) is obvious, by setting, in (1.2), βn=0\beta_{n}=0, for all nn\in\mathbb{N}. We will prove the implication (i) \Rightarrow (ii). Suppose that limnun=x\lim_{n\rightarrow\infty}u_{n}=x^{*}. If

limnxnun=0\lim_{n\rightarrow\infty}\left\|x_{n}-u_{n}\right\|=0 (2.3)

then

0xxnunx+xnun0\leq\left\|x^{*}-x_{n}\right\|\leq\left\|u_{n}-x^{*}\right\|+\left\|x_{n}-u_{n}\right\| (2.4)

and it follows that

limnxn=x\lim_{n\rightarrow\infty}x_{n}=x^{*} (2.5)

Thus, to complete the proof it suffices to verify relation (2.3).

With A:=(IT)A:=(I-T) in (1.9), we have

AxAy,J(xy)\displaystyle\langle Ax-Ay,J(x-y)\rangle =(xTx)(yTy),J(xy)\displaystyle=\langle(x-Tx)-(y-Ty),J(x-y)\rangle
=xy2TxTy,J(xy)\displaystyle=\|x-y\|^{2}-\langle Tx-Ty,J(x-y)\rangle
xy2xy2+ψ(xy)\displaystyle\geq\|x-y\|^{2}-\|x-y\|^{2}+\psi(\|x-y\|)
=ψ(xy)\displaystyle=\psi(\|x-y\|) (2.6)

Taking x:=xnx:=x_{n} and y:=uny:=u_{n} in (2.6) we obtain

AxnAun,J(xnun)ψ(xnun).\left\langle Ax_{n}-Au_{n},J\left(x_{n}-u_{n}\right)\right\rangle\geq\psi\left(\left\|x_{n}-u_{n}\right\|\right). (2.7)

Choose R>0R>0 such that {un:n}BR(x)\left\{u_{n}:n\in\mathbb{N}\right\}\subset B_{R}\left(x^{*}\right) and x0B2R(x)x_{0}\in B_{2R}\left(x^{*}\right). Remark 1.8 assures that A(B2R(x))A\left(B_{2R}\left(x^{*}\right)\right) is bounded. Denote

σ:=diam(A(B2R(x)))+R\sigma:=\operatorname{diam}\left(A\left(B_{2R}\left(x^{*}\right)\right)\right)+R (2.8)

Since the map J()J(\cdot) is uniformly continuous on bounded subsets of XX, with

ε:=ψ(R2)4σ>0\varepsilon:=\frac{\psi\left(\frac{R}{2}\right)}{4\sigma}>0 (2.9)

there exists a δ1>0\delta_{1}>0 such that xyδ1\|x-y\|\leq\delta_{1} implies J(x)J(y)ε\|J(x)-J(y)\|\leq\varepsilon.
The map T()T(\cdot) is also uniformly continuous. Thus for the same ε\varepsilon, there exits a δ2>0\delta_{2}>0 such that xyδ2\|x-y\|\leq\delta_{2} implies TxTyε\|Tx-Ty\|\leq\varepsilon.

We shall prove by induction that {xn}\left\{x_{n}\right\} is bounded. We know that 0=x0u0R0=\left\|x_{0}-u_{0}\right\|\leq R. Suppose that xkukR,k{1,,n}\left\|x_{k}-u_{k}\right\|\leq R,\forall k\in\{1,\ldots,n\}. We shall prove that

xn+1un+1R\left\|x_{n+1}-u_{n+1}\right\|\leq R (2.10)

Assume that xnunR\left\|x_{n}-u_{n}\right\|\leq R and that

xn+1un+1>R\left\|x_{n+1}-u_{n+1}\right\|>R (2.11)

From xkukR,k{1,,n}\left\|x_{k}-u_{k}\right\|\leq R,\forall k\in\{1,\ldots,n\} we know

xkxxkuk+ukx2R,k{1,,n}\left\|x_{k}-x^{*}\right\|\leq\left\|x_{k}-u_{k}\right\|+\left\|u_{k}-x^{*}\right\|\leq 2R,\forall k\in\{1,\ldots,n\} (2.12)

From (2.12), we have xnB2R(x)x_{n}\in B_{2R}\left(x^{*}\right) and the following inequality satisfied

xkxkx+x2R+x,k{1,,n}\left\|x_{k}\right\|\leq\left\|x_{k}-x^{*}\right\|+\left\|x^{*}\right\|\leq 2R+\left\|x^{*}\right\|,\forall k\in\{1,\ldots,n\} (2.13)

Usingdiam(A(B2R(x)))σ\operatorname{Using}\operatorname{diam}\left(A\left(B_{2R}\left(x^{*}\right)\right)\right)\leq\sigma and xnB2R(x)x_{n}\in B_{2R}\left(x^{*}\right), (i. e. Axnσ\left\|Ax_{n}\right\|\leq\sigma ), we get

TynTxn\displaystyle\left\|Ty_{n}-Tx_{n}\right\| yn+Tyn+xnTxn+ynxn\displaystyle\leq\left\|-y_{n}+Ty_{n}+x_{n}-Tx_{n}\right\|+\left\|y_{n}-x_{n}\right\|
=AynAxn+ynxn\displaystyle=\left\|Ay_{n}-Ax_{n}\right\|+\left\|y_{n}-x_{n}\right\|
Ayn+Axn+ynxn\displaystyle\leq\left\|Ay_{n}\right\|+\left\|Ax_{n}\right\|+\left\|y_{n}-x_{n}\right\|
S+σ+βnxnTxn=S+σ+βnAxn\displaystyle\leq S+\sigma+\beta_{n}\left\|x_{n}-Tx_{n}\right\|=S+\sigma+\beta_{n}\left\|Ax_{n}\right\|
S+σ+βnσ\displaystyle\leq S+\sigma+\beta_{n}\sigma (2.14)

Such a S>0S>0 exists because

yk\displaystyle\left\|y_{k}\right\| xk+βkAxkxk+Axk\displaystyle\leq\left\|x_{k}\right\|+\beta_{k}\left\|Ax_{k}\right\|\leq\left\|x_{k}\right\|+\left\|Ax_{k}\right\| (2.15)
2R+x+σ,k{1,,n}\displaystyle\leq 2R+\left\|x^{*}\right\|+\sigma,\forall k\in\{1,\ldots,n\}

and AA is a bounded map.
For all nn\in\mathbb{N}, we have

un+1un=αn(IT)un=αnAunαnσ.\left\|u_{n+1}-u_{n}\right\|=\alpha_{n}\left\|(I-T)u_{n}\right\|=\alpha_{n}\left\|Au_{n}\right\|\leq\alpha_{n}\sigma. (2.16)

Set

δ:=min{δ1,δ2}\delta:=\min\left\{\delta_{1},\delta_{2}\right\} (2.17)

Defining

d0:=min{1,δ,δ2σ,R2(4σ+S)}.d_{0}:=\min\left\{1,\delta,\frac{\delta}{2\sigma},\frac{R}{2(4\sigma+S)}\right\}. (2.18)

it follows that, for all nn\in\mathbb{N}, using (2.1) and (2.16), that

αn(3σ+S+βnσ)\displaystyle\alpha_{n}\left(3\sigma+S+\beta_{n}\sigma\right) αn(4σ+S)<R2,\displaystyle\leq\alpha_{n}(4\sigma+S)<\frac{R}{2},
βn\displaystyle\beta_{n} <δσ and\displaystyle<\frac{\delta}{\sigma}\text{ and }
αn\displaystyle\alpha_{n} <δ2σ.\displaystyle<\frac{\delta}{2\sigma}. (2.19)

From (1.1) and (1.2),

xn+1un+1\displaystyle\left\|x_{n+1}-u_{n+1}\right\| =(1αn)(xnun)+αn(TynTun)\displaystyle=\left\|\left(1-\alpha_{n}\right)\left(x_{n}-u_{n}\right)+\alpha_{n}\left(Ty_{n}-Tu_{n}\right)\right\|
=xnunαn(AxnAun)+αn(TynTxn)\displaystyle=\left\|x_{n}-u_{n}-\alpha_{n}\left(Ax_{n}-Au_{n}\right)+\alpha_{n}\left(Ty_{n}-Tx_{n}\right)\right\|
xnun+αnAxnAun+αnTynTxn.\displaystyle\leq\left\|x_{n}-u_{n}\right\|+\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|+\alpha_{n}\left\|Ty_{n}-Tx_{n}\right\|. (2.20)

From (2.20), using (2.11), (2.8), (2.14) and the first evaluation from (2.19),

xnun\displaystyle\left\|x_{n}-u_{n}\right\| xn+1un+1αnAxnAunαnTynTxn\displaystyle\geq\left\|x_{n+1}-u_{n+1}\right\|-\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|-\alpha_{n}\left\|Ty_{n}-Tx_{n}\right\|
R2αnσαn(S+σ+βnσ)\displaystyle\geq R-2\alpha_{n}\sigma-\alpha_{n}\left(S+\sigma+\beta_{n}\sigma\right)
=Rαn(3σ+S+βnσ)RR/2=R/2\displaystyle=R-\alpha_{n}\left(3\sigma+S+\beta_{n}\sigma\right)\geq R-R/2=R/2 (2.21)

Using the induction assumption,

xn+1un+1\displaystyle\left\|x_{n+1}-u_{n+1}\right\| =(1αn)(xnun)+αn(TynTun)\displaystyle=\left\|\left(1-\alpha_{n}\right)\left(x_{n}-u_{n}\right)+\alpha_{n}\left(Ty_{n}-Tu_{n}\right)\right\|
=(xnun)αn(xnunTxn+Tun)+αn(TynTxn)\displaystyle=\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(x_{n}-u_{n}-Tx_{n}+Tu_{n}\right)+\alpha_{n}\left(Ty_{n}-Tx_{n}\right)\right\|
xnun+αnAxnAun+αnTynTxn\displaystyle\leq\left\|x_{n}-u_{n}\right\|+\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|+\alpha_{n}\left\|Ty_{n}-Tx_{n}\right\|
R+2αnσ+αnS+αnσ+αnβnσ=R+αnS+3αnσ+αnβnσ\displaystyle\leq R+2\alpha_{n}\sigma+\alpha_{n}S+\alpha_{n}\sigma+\alpha_{n}\beta_{n}\sigma=R+\alpha_{n}S+3\alpha_{n}\sigma+\alpha_{n}\beta_{n}\sigma
<R+R/22R.\displaystyle<R+R/2\leq 2R. (2.22)

Thus we get

1xn+1un+12R.-1\leq-\frac{\left\|x_{n+1}-u_{n+1}\right\|}{2R}. (2.23)

By setting (1.6),

x\displaystyle x :=(xnun)αn(AxnAun),\displaystyle:=\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right),
y\displaystyle y :=αn(TynTxn),\displaystyle:=\alpha_{n}\left(Ty_{n}-Tx_{n}\right),
x+y\displaystyle x+y =xn+1un+1,\displaystyle=x_{n+1}-u_{n+1}, (2.24)

we obtain

xn+1un+12\displaystyle\left\|x_{n+1}-u_{n+1}\right\|^{2} =(1αn)(xnun)+αn(TynTun)2\displaystyle=\left\|\left(1-\alpha_{n}\right)\left(x_{n}-u_{n}\right)+\alpha_{n}\left(Ty_{n}-Tu_{n}\right)\right\|^{2}
=(xnun)αn(xnun)+αn(TxnTun)+αn(TynTxn)2\displaystyle=\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(x_{n}-u_{n}\right)+\alpha_{n}\left(Tx_{n}-Tu_{n}\right)+\alpha_{n}\left(Ty_{n}-Tx_{n}\right)\right\|^{2}
=(xnun)αn(AxnAun)+αn(TynTxn)2\displaystyle=\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)+\alpha_{n}\left(Ty_{n}-Tx_{n}\right)\right\|^{2}
(xnun)αn(AxnAun)2+2αnTynTxn,J(xn+1un+1).\displaystyle\leq\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right\|^{2}+2\alpha_{n}\left\langle Ty_{n}-Tx_{n},J\left(x_{n+1}-u_{n+1}\right)\right\rangle. (2.25)

We again apply (1.6) with

x\displaystyle x :=xnun,\displaystyle:=x_{n}-u_{n},
y\displaystyle y :=αn(AxnAun),\displaystyle:=-\alpha_{n}\left(Ax_{n}-Au_{n}\right),
x+y\displaystyle x+y =(xnun)αn(AxnAun),\displaystyle=\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right), (2.26)

to obtain,

(xnun)αn(AxnAun)2\displaystyle\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right\|^{2}
xnun22αnAxnAun,J((xnun)αn(AxnAun))\displaystyle\leq\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\left\langle Ax_{n}-Au_{n},J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)\right\rangle
xnun22αnAxnAun,J((xnun)αn(AxnAun))J(xnun)\displaystyle\leq\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\left\langle Ax_{n}-Au_{n},J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\rangle
2αnAxnAun,J(xnun)\displaystyle-2\alpha_{n}\left\langle Ax_{n}-Au_{n},J\left(x_{n}-u_{n}\right)\right\rangle
xnun22αnψ(xnun)\displaystyle\leq\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\left\|x_{n}-u_{n}\right\|\right)
+2αnAxnAun×J((xnun)αn(AxnAun))J(xnun).\displaystyle\quad+2\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|\times\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|. (2.27)

Substituting (2.27) into (2.25) and using (2.21) we have

xn+1un+12\displaystyle\left\|x_{n+1}-u_{n+1}\right\|^{2}
xnun22αnψ(xnun)\displaystyle\leq\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\left\|x_{n}-u_{n}\right\|\right)
+2αnAxnAun×J((xnun)αn(AxnAun))J(xnun)\displaystyle+2\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|\times\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|
+2αnTynTxn,J(xn+1un+1)\displaystyle+2\alpha_{n}\left\langle Ty_{n}-Tx_{n},J\left(x_{n+1}-u_{n+1}\right)\right\rangle
xnun22αnψ(xnun)\displaystyle\leq\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\left\|x_{n}-u_{n}\right\|\right)
+2αnAxnAun×J((xnun)αn(AxnAun))J(xnun)\displaystyle+2\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|\times\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|
+2αnTynTxnxn+1un+1\displaystyle+2\alpha_{n}\left\|Ty_{n}-Tx_{n}\right\|\left\|x_{n+1}-u_{n+1}\right\|
\displaystyle\leq xnun22αnψ(R2)\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\frac{R}{2}\right)
+2αnAxnAun×J((xnun)αn(AxnAun))J(xnun)\displaystyle+2\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|\times\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|
+2αnTynTxnxn+1un+1\displaystyle+2\alpha_{n}\left\|Ty_{n}-Tx_{n}\right\|\left\|x_{n+1}-u_{n+1}\right\|
\displaystyle\leq xnun22αnψ(R2)+4αnστn+2αnζnxn+1un+1.\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\frac{R}{2}\right)+4\alpha_{n}\sigma\tau_{n}+2\alpha_{n}\zeta_{n}\left\|x_{n+1}-u_{n+1}\right\|. (2.28)

Setting

τn:=J((xnun)αn(AxnAun))J(xnun)\tau_{n}:=\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\| (2.29)

and

ζn:=TynTxn,\zeta_{n}:=\left\|Ty_{n}-Tx_{n}\right\|, (2.30)

and using (2.8) and (2.23),

xn+1un+12\displaystyle\left\|x_{n+1}-u_{n+1}\right\|^{2}
xnun22αnψ(R2)xn+1un+12R+4αnστn+2αnζnxn+1un+1\displaystyle\leq\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{2R}+4\alpha_{n}\sigma\tau_{n}+2\alpha_{n}\zeta_{n}\left\|x_{n+1}-u_{n+1}\right\| (2.31)

Using (2.14) and (2.19) we obtain

(xnun)αn(AxnAun)(xnun)\displaystyle\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)-\left(x_{n}-u_{n}\right)\right\| (2.32)
=αn(AxnAun)2αnσ<δ.\displaystyle=\left\|\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right\|\leq 2\alpha_{n}\sigma<\delta.

From the uniform continuity of J()J(\cdot),

τnε.\tau_{n}\leq\varepsilon. (2.33)

Relation (2.19) leads to

ynxn=βnxn+βnTxn=βnAxnβnσ<δ\left\|y_{n}-x_{n}\right\|=\left\|-\beta_{n}x_{n}+\beta_{n}Tx_{n}\right\|=\beta_{n}\left\|Ax_{n}\right\|\leq\beta_{n}\sigma<\delta (2.34)

Since TT is uniformly continuous,

ζn<ε\zeta_{n}<\varepsilon (2.35)

Substituting (2.33), (2.35) (with ε\varepsilon given by (2.9)), and (2.23) in (2.31) we obtain

xn+1un+12\displaystyle\left\|x_{n+1}-u_{n+1}\right\|^{2}
\displaystyle\leq xnun22αnψ(R2)xn+1un+12R+4αnσψ(R2)4σ+2αnζnxn+1un+1\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{2R}+4\alpha_{n}\sigma\frac{\psi\left(\frac{R}{2}\right)}{4\sigma}+2\alpha_{n}\zeta_{n}\left\|x_{n+1}-u_{n+1}\right\|
\displaystyle\leq xnun2αnψ(R2)xn+1un+1R+αnψ(R2)+12αnψ(R2)σxn+1un+1\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{R}+\alpha_{n}\psi\left(\frac{R}{2}\right)+\frac{1}{2}\alpha_{n}\frac{\psi\left(\frac{R}{2}\right)}{\sigma}\left\|x_{n+1}-u_{n+1}\right\|
\displaystyle\leq xnun2αnψ(R2)xn+1un+1R+αnψ(R2)xn+1un+12R+\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{R}+\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{2R}+
+12αnψ(R2)xn+1un+1σ\displaystyle+\frac{1}{2}\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{\sigma}
=\displaystyle= xnun2αnψ(R2)xn+1un+1R+12αnψ(R2)xn+1un+1R+\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{R}+\frac{1}{2}\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{R}+
+12αnψ(R2)xn+1un+1R\displaystyle+\frac{1}{2}\alpha_{n}\psi\left(\frac{R}{2}\right)\frac{\left\|x_{n+1}-u_{n+1}\right\|}{R}
=\displaystyle= xnun2R2.\displaystyle\left\|x_{n}-u_{n}\right\|^{2}\leq R^{2}. (2.36)

Relation (2.36) is in contradiction with xn+1un+1>R\left\|x_{n+1}-u_{n+1}\right\|>R.
Thus there exists an R>0R>0 such that

xnunR,n\left\|x_{n}-u_{n}\right\|\leq R,\forall n\in\mathbb{N} (2.37)

Relations (2.28) and (2.37) lead to

xn+1un+12\displaystyle\left\|x_{n+1}-u_{n+1}\right\|^{2}\leq xnun22αnψ(xnun)\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\left\|x_{n}-u_{n}\right\|\right)
+2αnAxnAun×J((xnun)αn(AxnAun))J(xnun)\displaystyle+2\alpha_{n}\left\|Ax_{n}-Au_{n}\right\|\times\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|
+2αnTynTxnxn+1un+1\displaystyle+2\alpha_{n}\left\|Ty_{n}-Tx_{n}\right\|\left\|x_{n+1}-u_{n+1}\right\|
\displaystyle\leq xnun22αnψ(xnun)\displaystyle\left\|x_{n}-u_{n}\right\|^{2}-2\alpha_{n}\psi\left(\left\|x_{n}-u_{n}\right\|\right)
+4αnσJ((xnun)αn(AxnAun))J(xnun)\displaystyle+4\alpha_{n}\sigma\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|
+2αnRTynTxn\displaystyle+2\alpha_{n}R\left\|Ty_{n}-Tx_{n}\right\| (2.38)

Recalling that limnunx=0\lim_{n\rightarrow\infty}\left\|u_{n}-x^{*}\right\|=0, then limnun+1un=0\lim_{n\rightarrow\infty}\left\|u_{n+1}-u_{n}\right\|=0, and using (2.32) one obtains using (1.3),

(xnun)αn(AxnAun)(xnun)\displaystyle\left\|\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)-\left(x_{n}-u_{n}\right)\right\| (2.39)
=αn(AxnAun)2αnσ0 as n.\displaystyle=\left\|\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right\|\leq 2\alpha_{n}\sigma\rightarrow 0\text{ as }n\rightarrow\infty.

The uniformly continuity of J()J(\cdot) implies that

limnJ((xnun)αn(AxnAun))J(xnun)=0\lim_{n\rightarrow\infty}\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|=0 (2.40)

Also, from (2.34) and (1.3), we have

ynxn\displaystyle\left\|y_{n}-x_{n}\right\| =βnxn+βnTxn\displaystyle=\left\|-\beta_{n}x_{n}+\beta_{n}Tx_{n}\right\| (2.41)
=βnAxnβnσ0, as n.\displaystyle=\beta_{n}\left\|Ax_{n}\right\|\leq\beta_{n}\sigma\rightarrow 0,\text{ as }n\rightarrow\infty.

The uniformly continuity of T()T(\cdot) leads to

limnTynTxn=0\lim_{n\rightarrow\infty}\left\|Ty_{n}-Tx_{n}\right\|=0 (2.42)

Relations (2.38), (2.40) and (2.42) with

λn\displaystyle\lambda_{n} :=xnun2\displaystyle:=\left\|x_{n}-u_{n}\right\|^{2} (2.43)
γn\displaystyle\gamma_{n} :=αn(4σJ((xnun)αn(AxnAun))J(xnun)\displaystyle:=\alpha_{n}\left(4\sigma\left\|J\left(\left(x_{n}-u_{n}\right)-\alpha_{n}\left(Ax_{n}-Au_{n}\right)\right)-J\left(x_{n}-u_{n}\right)\right\|\right.
+2RTynTxn)\displaystyle\left.+2R\left\|Ty_{n}-Tx_{n}\right\|\right)

lead to (1.12). Using now Lemma 1.9 one obtains limnxnun2=0\lim_{n\rightarrow\infty}\left\|x_{n}-u_{n}\right\|^{2}=0.
Using Remark 1.4 (i), Proposition 1.5, and Theorem 2.1 one obtains the following corollary.

Corollary 2.2. Let XX be a real Banach space with XX^{\prime} strictly convex. If T:XXT:X\rightarrow X is a dd-weakly contractive (respectively weakly contractive) and uniformly continuous map with xF(T),x0=u0Xx^{*}\in F(T),x_{0}=u_{0}\in X and there exists a constant d0=d0(T,x)(0,1)d_{0}=d_{0}\left(T,x^{*}\right)\in(0,1), which depends on TT and xx^{*}, such that {αn},{βn}\left\{\alpha_{n}\right\},\left\{\beta_{n}\right\} satisfy αn,βnd0,n\alpha_{n},\beta_{n}\leq d_{0},\forall n\in\mathbb{N} and (1.3), then the following are equivalent:
(i) the Mann iteration (1.1) converges to the xF(T)x^{*}\in F(T),
(ii) the Ishikawa iteration (1.2) converges to the same xx^{*}.

Let CC be a ψ\psi-uniformly accretive map. Suppose the equation Cx=fCx=f has a solution for a given ff. Remark 1.4 (ii) ensures that

Tx:=f+xCx,xX,Tx:=f+x-Cx,\forall x\in X, (2.44)

is a ψ\psi-uniformly pseudocontractive map. A fixed point for TT is a solution for Cx=fCx=f and conversely.

Theorem 2.1 also implies the following corollary.
Corollary 2.3. Let XX be a real Banach space with XX^{\prime} strictly convex. If C:XXC:X\rightarrow X is a ψ\psi-uniformly accretive and uniformly continuous map with xF(T),x0=u0Xx^{*}\in F(T),x_{0}=u_{0}\in X and there exists a constant d0=d0(T,x)(0,1)d_{0}=d_{0}\left(T,x^{*}\right)\in(0,1), which depends on TT and xx^{*}, such that {αn},{βn}\left\{\alpha_{n}\right\},\left\{\beta_{n}\right\} satisfy αn,βnd0,n\alpha_{n},\beta_{n}\leq d_{0},\forall n\in\mathbb{N} and (1.3), then the following are equivalent:
(i) the Mann iteration (1.1), with TT given by (2.44), converges to the solution of Cx=fCx=f,
(ii) the Ishikawa iteration (1.2), with TT given by (2.44), converges to the solution of Cx=fCx=f.

References

[1] Ya. I. Alber and S. Reich, An iterative method for solving a class of nonlinear operator equations in Banach spaces, Panamer. Math. J. 4(1994), 39-54.
[2] Ya. I. Alber and S. Guerre-Delabriere, Principle of weakly contractive maps in Hilbert spaces, Operator Theory 98 (1997), 7-22.
[3] Ya. I. Alber and S. Guerre-Delabriere, On the projection methods for fixed point problems, Analysis 21(2001), 17-39.
[4] M. S. Berger, Nonlinearity and Functional Analysis, Academic Press, New York, 1977, page 65.
[5] C. E. Chidume, H. Zegeye and S. J. Aneke, Approximation of fixed points of weakly contractive nonself maps in Banach spaces, J. Math. Anal. Appl. 270(2002), 189-199.
[6] C. E. Chidume and H. Zegeye, Approximation methods for nonlinear operator equations, Proc. Amer. Math. Soc. 131(2003), 2467-2478.
[7] K. Deimling, Nonlinear Functional Analysis, Springer-Verlag, 1985, page 115.
[8] S. Ishikawa, Fixed points by a new iteration method, Proc. Amer. Math. Soc. 44 (1974), 147-150.
[9] W. R. Mann, Mean value in iteration, Proc. Amer. Math. Soc. 4(1953), 506-510.
[10] C. Morales and J. S. Jung, Convergence of paths for pseudocontractive mappings in Banach spaces, Proc. Amer. Math. Soc. 128(2000), 3411-3419.
[11] B. E. Rhoades and Ş. M. Şoltuz, On the equivalence of Mann and Ishikawa iteration methods, Int. J. Math. Math. Sci. 2003(7), 451-459.
[12] B. E. Rhoades and Ş. M. Şoltuz, The equivalence of Mann and Ishikawa iteration for ψ\psi-uniformly pseudocontractive or ψ\psi-uniformly accretive maps, Int. J. Math. Math. Sci. 46(2004), 2443-2452.

Department of Mathematics, Indiana University, Bloomington, IN 47405-7106, U.S.A.
E-mail: rhoades@indiana.edu
Str. Avram Iancu 13, Ap. 1, 400083 Cluj-Napoca, Romania
E-mail: soltuzul@yahoo.com smsoltuz@gmail.com

2006

Related Posts