• 検索結果がありません。

Approximate Fixed Points for Nonexpansive and Quasi-Nonexpansive Mappings in Hyperspaces

N/A
N/A
Protected

Academic year: 2022

シェア "Approximate Fixed Points for Nonexpansive and Quasi-Nonexpansive Mappings in Hyperspaces"

Copied!
16
0
0

読み込み中.... (全文を見る)

全文

(1)

Volume 2009, Article ID 520976,16pages doi:10.1155/2009/520976

Research Article

Approximate Fixed Points for Nonexpansive and Quasi-Nonexpansive Mappings in Hyperspaces

Zeqing Liu,

1

Jeong Sheok Ume,

2

and Shin Min Kang

3

1Department of Mathematics, Liaoning Normal University, Dalian, Liaoning 116029, China

2Department of Applied Mathematics, Changwon National University, Changwon 641-773, South Korea

3Department of Mathematics, Research Institute of Natural Science, Gyeongsang National University, Chinju 660-701, South Korea

Correspondence should be addressed to Jeong Sheok Ume,[email protected] Received 9 May 2009; Accepted 14 December 2009

Recommended by W. A. Kirk

This paper provides a few convergence results of the Ishikawa iteration sequence with errors for nonexpansive and quasi-nonexpansive mappings in hyperspaces. The results presented in this paper improve and generalize some results in the literature.

Copyrightq2009 Zeqing Liu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction and Preliminaries

Browder 1 and Kirk 2 established that a nonexpansive mapping T which maps a closed bounded convex subset C of a uniformly convex Banach space into itself has a fixed point inC. Since then, many researchers have studied, under various conditions, the convergence of the Mann and Ishikawa iteration methods dealing with nonexpansive and quasi-nonexpansive mappingssee3–11and the references therein. Rhoades9pointed out that the Picard iteration schemes for nonexpansive mappings need not converge. Senter and Dotson 10 obtained conditions under which the Mann iteration schemes generated by nonexpansive and quasi-nonexpansiv mappings in uniformly convex Banach spaces, converge to fixed points of these mappings, respectively. Ishikawa7established that the Mann iteration methods can be used to approximate fixed points of nonexpansive mappings in Banach spaces. Deng 3 obtained similar results for Ishikawa iteration processes in normed linear spaces and Banach spaces.

Our aim is to prove several convergence theorems of the Ishikawa iteration sequence with errors for nonexpansive and quasi-nonexpansive mappings in hyperspaces. Our results presented in this paper extend substantially the results due to Deng3, Ishikawa7, and Senter and Dotson10.

(2)

Assume thatX is a nonempty subset of a normed linear spaceE, · and CCX denotes the family of all nonempty convex compact subsets ofX, and H is the Hausdorff metric induced by the norm · . ForxE,XE,A, BCCX,I⊆CCX,T :I, H → CCX, H, andtR −∞,∞, let

dx, , A inf{x−a:aA}, DA,I inf{HA, C:C∈I}, IX {{x}:xX}, AB{ab:aA, bB}, tA{ta:aA}, coI

n

i1

tiAi:ti≥0, n

i1

ti1, Ai∈I, n≥1

, FT {A∈I:TAA}.

1.1

It is easy to see that tA 1−tA Aand tA 1−tBCCE for allt ∈ 0,1and A, BCCE. HenceCCEis convex. Hu and Huang12proved that ifE, · is a Banach space, thenCCX, His a complete metric space. Now we introduce the following concepts in hyperspaces.

Definition 1.1. LetIbe a nonempty subset ofCCEand letT : I, H → CCE, Hbe a mapping. Assume that{tn}n≥0,{tn}n≥0,{sn}n≥0, and{sn}n≥0 are arbitrary real sequences in 0,1satisfyingtntn ≤1 andsnsn≤1 forn≥1 and{Pn}n≥0and{Qn}n≥0are any bounded sequences of the elements inCCE.

iForA0∈I, the sequence{An}n≥0defined by Bn

1−snsn

AnsnTAnsnPn, An1

1−tntn

AntnTBntnQn, n≥0 1.2 is called the Ishikawa iteration sequence with errors provided that{An, Bn : n ≥ 0} ⊆I.

iiIfsntn0 for alln≥0 in1.2, the sequence{An}n≥0defined by

Bn 1−snAnsnTAn, An1 1−tnAntnTBn, n≥0, 1.3 is called the Ishikawa iteration sequence provided that{An, Bn:n≥0} ⊆I.

iiiIfsnsn0 for alln≥0 in1.2, the sequence{An}n≥0defined by An1

1−tntn

AntnTAntnQn, n≥0, 1.4 is called the Mann iteration sequence with errors provided that{An:n≥0} ⊆I.

ivIfsntnsn0 for alln≥0 in1.2, the sequence{An}n≥0defined by

An1 1−tnAntnTAn, n≥0, 1.5 is called the Mann iteration sequence provided that{An:n≥0} ⊆I.

(3)

Definition 1.2. LetIbe a nonempty subset ofCCE. A mappingT :I, H → CCE, His said to be

inonexpansive ifHTA, TBHA, Bfor allA, B∈I;

iiquasi-nonexpansive ifFT/∅andHTA, PHA, Pfor allA∈IandPFT. Definition 1.3. LetIbe a nonempty subset ofCCE. A mappingT :I, H → CCE, H withFT/∅is said to be satisfy the following.

iCondition A if there is a continuous functionf:0,∞ → 0,∞withf0 0 and ft>0 fort∈0,∞, such thatHA, TAfDA, FTfor allA∈I.

iiCondition B if there is a nondecreasing functionf :0,∞ → 0,∞withf0 0 andft>0 fort∈0,∞, such thatHA, TAfDA, FTfor allA∈I.

Remark 1.4. In caseIIX, whereXis a nonempty subset ofE, andT :IX → IECCEis a mapping, then Definitions1.1,1.2, and1.3iireduce to the corresponding concepts in1–

11,13. It is well known that every nonexpansive mapping with nonempty fixed point set is quasi-nonexpansive, but the converse is not true; see8. Examples3.1and3.4in this paper reveal that the class of nonexpansive mappings with nonempty fixed point set is a proper subclass of quasi-nonexpansive mappings with both Condition A and Condition B.

The following lemmas play important roles in this paper.

Lemma 1.5see12. LetE, · be a Banach space andIa compact subset ofCCE, H. Then coI, His compact, wherecoIstands for the closure ofcoI.

Lemma 1.6see4. Suppose that{an}n≥0,{bn}n≥0, and{cn}n≥0are three sequences of nonnegative numbers such that an1 ≤ 1 bnan cn for alln0. If

n0bn and

n0cn converge, then limn→ ∞anexists.

Lemma 1.7see14. LetX, dbe a metric space. LetAandBbe compact subsets ofX. Then for anyaA, there existsbBsuch thatda, bHA, B, whereHis the Hausdorffmetric induced byd.

Lemma 1.8. LetE, · be a normed linear space. Then

H1tsAtBsC,1−tsLtMsN≤1−tsHA, L tHB, M sHC, N 1.6 for allA, B, C, L, M, NCCEandt, s∈0,1withst1.

Proof. Set

r 1−tsHA, L tHB, M sHC, N. 1.7

For anyaA,bB,cC, byLemma 1.7we infer that there existlL,mM,nNsuch that

a−l ≤HA, L, b−m ≤HB, M, c−n ≤HC, N, 1.8

(4)

which imply that

1−tsatbsc−1−tsltmsn ≤1−tsaltbmscn ≤r.

1.9

That is,

sup{d1−tsatbsc,1−tsLtMsN:aA, bB, cC} ≤r. 1.10

Similarly we have

sup{d1−tsltmsn,1−tsAtBsC:lL, mM, nN} ≤r. 1.11

Thus1.6follows from1.10and1.11. This completes the proof.

Lemma 1.9. LetE, · be a normed linear space andIa nonempty closed subset ofCCE, H.

IfT :I, H → CCE, His quasi-nonexpansive, thenFTis closed.

Proof. Let {Pn}n≥0 be inFTwith limn→ ∞HPn, P 0. Since T is quasi-nonexpansive, it follows that

HP, TPHPn, P HPn, TP≤2HPn, P−→0 1.12

asn → ∞. HencePFT. That is,FTis closed. This completes the proof.

2. Main Results

Our results are as follows.

Theorem 2.1. LetE,·be a normed linear space and letIbe a nonempty subset ofCCE. Assume thatT :I, H → CCE, His nonexpansive andA0 ∈I. Suppose that there exists a constantt satisfying

0< tntnt <1, n≥0, 2.1

n0

tn∞,

n0

sn<∞,

n0

sn<∞,

n0

tn<∞,

n0

tn

tntn−1

<∞. 2.2

If the Ishikawa iteration sequence with errors{An}n≥0is bounded, then limn→ ∞HAn, TAn 0.

Proof. SinceT is nonexpansive,{An}n≥0,{Pn}n≥0, and{Qn}n≥0are bounded, it follows that a:sup{HA, B:A∈ {An, Bn, Pn, Qn:n≥0}, B∈ {An, Bn, TAn, TBn:n≥0}}<∞. 2.3

(5)

Let nand k be arbitrary nonnegative integers. In view of 1.2,2.3,Lemma 1.8, and the nonexpansiveness ofT, we conclude that

HBn, AnsnHAn, TAn asn, 2.4

HTBn, AnHTBn, TAn HTAn, An≤1snHAn, TAn asn, 2.5 HAn1, AntnHTBn, An atntn1snHAn, TAn a

tnsntn

, 2.6

HAn1, TAk

1−tntn

HAn, TAk tnHTBn, TAk atn

1−tntn

HAn, TAk tnHBn, Ak atn,

2.7

which yields that

HAn, TAk

1−tntn−1

HAn1, TAktnHBn, Akatn

. 2.8

Using1.2,2.3–2.6,Lemma 1.8, and the nonexpansiveness ofT, we have

HBn, Ank1HBn, An1

k i1

HAni, Ani1

1−snsn

HAn, An1 snHTAn, An1 asn

k

i1

1snitniHAni, TAni a

tnisnitni

1−snsn

tn1snHTAn, An a

tnsntn sn

1−tntn

HTAn, An tnHTBn, TAn atn asn

k

i1

1snitniHAni, TAni a k

i1

tnisnitni

tnsnsntns2ntnsntnsntnsntnsn

HAn, TAn

a

1−snsn

tnsntn sntn

snHAn, TAn asn

asntnasnk

i1

1snitniHAni, TAni a k

i1

tnisnitni

k

i0

tnisniHAni, TAni a

snk

i0

tnisnitni ,

2.9

(6)

HTAn1, An1

1−tntn

HAn, TAn1 tnHTBn, TAn1 tnHQn, TAn1

1−tntn

HAn1, TAn1 HAn, An1 tnHBn, An1 atn

1−tntn

HAn1, TAn1

1−tntn

1sntnHAn, TAn a

tnsntn tn

tnsnHAn, TAn a

sntnsntn atn

1−tntn

HAn1, TAn1 tn12snHAn, TAn 2a

tnsntn 2.10

which implies that

HAn1, TAn1

tntn−1

tn12snHAn, TAn 2a

tnsntn

≤12snHAn, TAn 2a sntn

tntn−1 .

2.11

Lemma 1.6,2.2, and2.11yield that there exists a nonnegative constantrsatisfying

nlim→ ∞HAn, TAn r, 2.12

which implies that for anyε >0 there exists a positive integerNsuch that

rεHAn, TAn fornN. 2.13

Now we prove by induction that the following inequality holds for alln≥1:

H

Ap, TApn

≥rε

1n−1

i0

tpi

−2ε n−1

i0

1−tpitpi−1

−rεn−1

i0

tpi

n−1

ji

spj

i

k0

1−tpktpk−1

a n−1

i0

⎧⎨

tpi

spin−1

ji

tpjspjtpj

tpi

×i

k0

1−tpktpk−1

, pN.

2.14

(7)

According to1.2,2.8,2.9, and2.13, we derive that H

Ap, TAp1

1−tptp−1 H

Ap1, TAp1

tpH

Bp, Ap1

atp

1−tptp−1

rε−rεtp

tpsp

atp

sptpsptp

atp

1−tp−tp−1

rε−rε 1−2

1−tp

1−tp

2 tpsp

a tp

sptpsptp tp

≥rε 1tp

−2ε

1−tptp−1

−rεtpsp

1−tptp−1

a tp

sptpsptp tp

1−tptp−1

, pN.

2.15 Hence2.14holds forn1. Suppose that2.14holds fornm≥1. That is,

H

Ap, TApm

≥rε

1m−1

i0

tpi

−2ε m−1

i0

1−tpitpi−1

−rεm−1

i0

tpi

m−1

ji

spj

i

k0

1−tpktpk−1

a

m−1

i0

⎧⎨

tpi

spim−1

ji

tpjspjtpj

tpi

×i

k0

1−tpktpk−1

, pN.

2.16

In view of1.2,2.8,2.9, and2.16, we infer that H

Ap, TApm1

1−tptp−1 H

Ap1, TApm1

tpH

Bp, Apm1

atn

1−tptp−1 rε

1m−1

i0

tp1i

−2ε m−1

i0

1−tp1itp1i−1

−rεm−1

i0

tp1i

m−1

ji

sp1j

i

k0

1−tp1ktp1k−1

a

m−1

i0

tp1i

sp1im−1

ji

tp1jsp1jtp1j

tp1i

×i

k0

1−tp1ktp1k−1

tp m−1

i0

tpispi rε

−atp

spm

i0

tpispitpi

atp

(8)

−2εm

i0

1−tpitpi−1

1−tptp−1 rε

×

1m

i0

tp1i−12 1−tp

− 1−tp

2tp

m i1

tpitp

m i0

spi

−rεm−1

i0

tp1i

m−1

ji

sp1j

i1

k0

1−tpktpk−1

a

tp

spm

i0

tpispitpi tp

1−tptp−1

m−1

i0

tp1i

sp1im−1

ji

tp1jsp1jtp1j

tp1i

×i1

k0

1−tpktpk−1

⎫⎬

⎭ −2εm

i0

1−tpitpi−1

rε 1−tp

1−tptp−1 1m

i0

tpi

−rε

⎧⎨

tp

1−tptp−1m

i0

spim−1

i0

tp1i

m−1

ji

sp1j

i1

k0

1−tpktpk−1

⎫⎬

a m

i0

⎧⎨

tpi

spim

ji

tpjspjtpj

tpi

i

k0

1−tpktpk−1

≥rε

1m

i0

tpi

−2ε m

i0

1−tpitpi−1

rεm

i0

tpi

m

ji

spj

i

k0

1−tpktpk−1

a m

i0

⎧⎨

tpi

spim

ji

tpjspjtpj

tpi

i

k0

1−tpktpk−1

, pN.

2.17

That is,2.14holds fornm1. Hence2.14holds for alln≥1.

We now assert thatr0. If not, thenr >0. Letmbe an arbitrary positive integer and

εmin

!

r,2−1rt2ra−11−tm, r1tm

2at−1−1"

. 2.18

(9)

According to2.1,2.2, and2.12, we know that there exists a positive integerN satisfying2.13and

max np

kn

sk, sinp

kn

tksktk

< ε forn, iN, p≥1. 2.19

It follows from2.1,2.2,2.13,2.14, and2.19that

HAN, TANm≥rε

1m−1

i0

tNi

−2ε m−1

i0

1−tNitNi−1

−rεm−1

i0

tNi

m−1

ji

sNj

i

k0

1−tNktNk−1

a

m−1

i0

⎧⎨

tNi

sNim−1

ji

tNjsNjtNj

tNi

×i

k0

1−tNktNk−1

≥rε

1m−1

i0

tNi

−2ε1−t−m

−rεεm−1

i0

tNi1−t−i−1a

m−1

i0

tNiεtNi

1−t−i−1

≥rε

1m−1

i0

tNi

−2ε1−t−m−rεεm−1

i0

tNi

i j0

1−t−j−1

m−1

i0

tNi i j0

1−t−j−1a

m−1

i0

⎣1−t−i−1i

j0

tNj

≥rε

1m−1

i0

tNi

−2ε1−t−m−rεεt−11−t−mm−1

i0

tNi

aεt−11−t−mm−1

i0

tNi

m−1

i0

1−t−i−1

−rεaεt−11−t−mm−1

i0

tNi

−2ε1−t−maεt−11−t−m

r−2raεt−11−t−mm−1

i0

tNi r

2at−1

ε1t−m

≥2−1r

m−1

i0

tNi−→ ∞

2.20

(10)

asm → ∞. Thus2.3and 2.20 yield thata ∞, which is absurd. Hence r 0. This completes the proof.

Theorem 2.2. LetE,·be a Banach space andIa nonempty closed subset ofCCE. Assume that T :I, H → CCE, His nonexpansive and there exists a compact subsetΩofCCEsuch that TI∪ {Pn, Qn :n≥0} ⊆Ω.If 2.1and2.2hold, thenThas a fixed point inI. Moreover, given A0∈I, the Ishikawa iteration sequence with errors{An}n≥0converges to a fixed point ofT.

Proof. SettingI0 co{A0} ∪Ω, byLemma 1.5and the compactnessΩwe conclude thatI0

is compact. It is evident that{An}n≥0 ⊆ I0, which yields that{An}n≥0is bounded. SinceIis closed and{An}n≥0 ⊆I, we conclude that there exist a subsequence{Ani}i≥0of{An}n≥0and A∈Isuch that

ilim→ ∞HAni, A 0. 2.21

It follows from2.21,Theorem 2.1, and the nonexpansiveness ofTthat HA, TAHA, Ani HAni, TAni HTAni, TA

≤2HA, Ani HAni, TAni−→0 2.22

asi → ∞. That is,ATA. Put

bsup{HPn, A, HQn, A:n≥0}. 2.23

In view of1.2,Lemma 1.8and the nonexpansiveness ofT, we derive that HAn1, A

1−tntn

HAn, A tnHTBn, A btn

1−tntn

HAn, A tn

1−snsn

HAn, A snHTAn, A bsn btn

2.24

forn ≥ 0. It follows fromLemma 1.6,2.2,2.23, and2.24that limi→ ∞HAn, Aexists.

Using2.21we get that limi→ ∞HAn, A 0. This completes the proof.

Theorem 2.3. LetE, · be a Banach space andIa nonempty closed subset of CCE. Suppose thatT :I, H → CCE, His a qusi-nonexpansive mapping and satisfies Condition A. Assume that2.1and2.2hold andA0is inI. IfFTis bounded, then the Ishikawa iteration sequence with errors{An}n≥0converges to a fixed point ofT inI.

Proof. Let b sup{HPn, A, HQn, A : n ≥ 0 and AFT}. Thenb < ∞. As in the proof ofTheorem 2.2, we get that2.24holds and limi→ ∞HAn, Aexists, whereAFT. Consequently,{An}n≥0is bounded and

DAn1, FTDAn, FT b sntn

∀n≥0. 2.25

(11)

It follows fromLemma 1.6,2.2, and2.25that limn→ ∞DAn, FT s ≥ 0. In view of Theorem 2.1and Condition A, we have

n→ ∞limHAn, TAn 0, fDAn, FTHAn, TAn ∀n≥0. 2.26

Using the continuity off, we know thatfs 0. That is,s0 and

nlim→ ∞DAn, FT 0. 2.27

Clearly2.27ensures that for anyi≥0 there existNi≥1 andPiFTsuch thatHANi, Pi<

2−i,which implies from2.24that

HAn, Pi<2−ib n−1 kNi

sktk

fornNi. 2.28

We requireNi1> Nifor alli≥0. Notice that for anyj > i≥0

H Pi, Pj

j−1

ki

HPk, ANk1 HANk1, Pk1

j−1

ki

2−kb

Nk1−1 mNk

smtm 2−k−1

3

2−i−2−j b

Nj−1 lNi

sltl .

2.29

Thus 2.2 and 2.29 yield that {Pi}i≥0 is a Cauchy sequence in FT. It follows from Lemma 1.9that there existsPFTsatisfying limi→ ∞Pi P. For anyε > 0 there exists i0>0 such that

max

⎧⎨

⎩2−i0, HPi0, P, bn−1

kNi0

sktk

<3−1ε forn > Ni0. 2.30

Using2.28and2.30we have

HAn, PHAn, Pi0 HPi0, P

≤2−i0b n−1 kNi0

sktk

HPi0, P

< ε

2.31

forn > Ni0. That is,{An}n≥0converges toPFT. This completes the proof.

(12)

A proof similar to that ofTheorem 2.3gives the following result and is thus omitted.

Theorem 2.4. Let E, · be a Banach space and let I be a nonempty closed subset ofCCE.

Suppose thatT :I, H → CCE, His a qusi-nonexpansive mapping and satisfies Condition B.

Assume thatA0is inIand there exists a constanttsatisfying

0< tnt <1, n≥0, n0

tn∞,

n0

sn<∞. 2.32

Then the Ishikawa iteration sequence{An}n≥0converges to a fixed point ofT inI.

LetX be a nonempty subset ofE, · . It is easy to see thatIX, His isometric to X, · . Thus Theorems2.1–2.4yield the following results.

Corollary 2.5. LetX be a nonempty subset of a normed linear space E, · . Assume that T : X, · → E, · is nonexpansive andA0X. Suppose that2.1and2.2hold. If the Ishikawa iteration sequence with errors{An}n≥0is bounded, then limn→ ∞AnTAn0.

Remark 2.6. Corollary 2.5 extends Theorem 1 in3and Lemma 2 in 7from the Ishikawa iteration scheme and Mann iteration scheme into the Ishikawa iteration scheme with errors, respectively.

Corollary 2.7. LetX be a nonempty closed subset of a Banach space E, · . Assume that T : X, · → E, · is nonexpansive and there exists a compact subsetY ofEwithTX∪ {Pn, Qn: n ≥ 0} ⊆ Y. Suppose that2.1and 2.2 hold. ThenT has a fixed point inX. Moreover for any A0X, the Ishikawa iteration sequence with errors{An}n≥0converges to a fixed point ofT.

Remark 2.8. Theorem 3 in3and Theorem 1 in7and8are special cases ofCorollary 2.7.

Corollary 2.9. LetXbe a nonempty closed subset of a Banach spaceE, · and letT :X, · → E,·be quasi-nonexpansive. Assume that2.1and2.2hold andTsatisfies Condition A. IfFT is bounded, then for anyA0X, the Ishikawa iteration sequence with errors{An}n≥0converges to a fixed point ofT inX.

Corollary 2.10. LetXbe a nonempty closed subset of a Banach spaceE,·and letT :X,· → E, · be quasi-nonexpansive. Assume that2.32holds andA0is inX. IfT satisfies Condition B, then the Ishikawa iteration sequence{An}n≥0converges to a fixed point ofTinX.

Remark 2.11. Corollary 2.10extends, improves, and unifies Theorem 4 in3, Theorem 2 in7 and8in the following ways:

ithe Mann iteration method in 7, 8, and Ishikawa iteration method in 3 are replaced by the more general Ishikawa iteration method with errors;

iithe nonexpansive mappings in 3, 7, 8 are replaced by the more general quasi- nonexpansive mappings.

(13)

3. Examples and Problems

Now we construct a few nontrivial examples to illustrate the results in Section 2. The following example reveals thatCorollary 2.10extends properly Theorem 4 in3, Theorem 2 in7and8.

Example 3.1. LetE Rwith the usual norm| · | and letX 0,1. DefineT : XEand f:0,∞ → 0,∞by

Tx

⎧⎪

⎪⎪

⎪⎪

⎪⎩ 3

4x, forx

$ 0,1

2

% , 1

2x, forx

&

1 2,1

% ,

3.1

andft 1/4tfort≥0. Settn 2√

n−1andsn 1n2−1forn≥0 andA0X. Then FT {0}and

|Tx−0| ≤ 3

4|x−0|, |x−Tx| ≥1

4|x|fdx, FT for xX. 3.2

Thus the assumptions ofCorollary 2.10are satisfied. However, Theorem 4 in3, Theorem 2 in7and8are not applicable since

''''T1 2 −T17

32 '''' 7

64 > 1 32 ''

''1 2− 17

32

'''', 3.3

that is,Tis not nonexpansive.

The examples below show that Theorems2.1–2.4extend substantially Corollaries2.5–

2.10, respectively.

Example 3.2. Let E R2 with the usual norm | · | and let X 0,12. For any a, b ∈ X, Δ0,0a,00, b stands for the triangle with vertices 0,0,a,0, and 0, b. Let I {Δ0,0a,00, b : a, b ∈ X}and {Pn}n≥0 and{Qn}n≥0 be inI. DefineT : I → CCE by

TΔ0,0a,00, b Δ0,0

2−1ab,0

0,4−1'''b2a2'''

fora, b∈X. 3.4

(14)

Puttn 2√

n−1,tn 2n7/4−1,sn sn 3n3/2−1forn≥0 andA0X. It follows thatIis a compact subset ofCCE,FT {Δ0,00,00,0}and

HTΔ0,0a,00, b, TΔ0,0c,00, d H

Δ0,0

2−1ab,0

0,4−1'''b2a2'''

,Δ0,0

2−1cd,0

0,4−1'''d2c2''' max(

2−1|abcd|,4−1)))'''b2a2'''−'''d2c2''')))*

≤max{|a−c|,|b−d|}

HΔ0,0a,00, b,Δ0,0c,00, d

3.5

fora, b,c, d ∈X. That is, the conditions of Theorems2.1and2.2are fulfilled. Hence we can invoke our Theorems2.1and2.2show that the Ishikawa iteration sequence with errors {An}n≥0converges toΔ0,00,00,0and limn→ ∞HAn, TAn 0.

Example 3.3. LetE, X,I,{Pn}n≥0,{Qn}n≥0,{sn}n≥0,{sn}n≥0,{tn}n≥0,{tn}n≥0, andA0 be as in Example 3.2. DefineT :I → CCEandf :0,∞ → 0,∞by

TΔ0,0a,00, b Δ0,0

&

a3

1a2−1 ,0

+&

0, b3

1b2−1+

fora, b∈X, ft t

1t2−1

fort∈0,∞.

3.6

Obviously,FT {Δ0,00,00,0},

HTΔ0,0a,00, b,Δ0,00,00,0 H

&

Δ0,0

&

a3

1a2−1 ,0

+&

0, b3

1b2−1+

,Δ0,00,00,0 +

max

! a3

1a2−1 , b3

1b2−1"

≤max{a, b}

HΔ0,0a,00, b,Δ0,00,00,0, HΔ0,0a,00, b, TΔ0,0a,0b,0

H

&

Δ0,0a,0b,0,Δ0,0

&

a3

1a2−1 ,0

+&

0, b3

1b2−1++

max,

fa, fb-

fmax{a, b}

fDΔ0,0a,00, b, FT

3.7

fora, b∈X. Therefore the conditions ofTheorem 2.3are fulfilled.

(15)

Example 3.4. LetE, X,I, andA0 be as inExample 3.2. DefineT :I → CCE,f : 0,∞ → 0,∞andh:0,1 → 1,2by

TΔ0,0a,00, b Δ0,0

2−1aha,0

0,2−1b2

fora, b∈X, ft 8−1t fort≥0,

hx

⎧⎪

⎪⎪

⎪⎪

⎪⎩ 7

4, forx

$ 0,1

2

% ,

1, forx

&

1 2,1

% .

3.8

It follows thatFT {Δ0,00,00,0},

HTΔ0,0a,00, b,Δ0,00,00,0 max(

2−1aha,2−1b2*

≤max{a, b}

HΔ0,0a,00, b,Δ0,00,00,0, HΔ0,0a,00, b, TΔ0,0a,00, b

H

Δ0,0a,0b,0,Δ0,0

2−1aha,0

0,2−1b2 max(

a

1−2−1ha , b

1−2−1b2*

≥8−1max{a, b}

fDΔ0,0a,00, b, FT

3.9

fora, b∈X. Obviously, the assumptions ofTheorem 2.4are fulfilled. On the other hand,T is not nonexpansive since

H

&

TΔ0,0

&

1 2,0

+&

0,1 2

+

, TΔ0,0

&

9 16,0

+&

0,1 2

++

1 2

''''1 2h

&

1 2

+

− 9 16h

&

9 16

+'''' 5

32 > 1 16 H

&

Δ0,0

&

1 2,0

+&

0,1 2

+

,Δ0,0

&

9 16,0

+&

0,1 2

++

.

3.10

We conclude with the following problems.

Problem 3.5. Can Condition A inTheorem 2.3be replaced by Condition B?

(16)

Problem 3.6. Can the boundedness ofFTinTheorem 2.3be removed?

Problem 3.7. CanTheorem 2.4be extended to the Ishikawa iteration method with errors?

Acknowledgment

This work was supported by the Korea Research Foundation Grant funded by the Korean GovernmentKRF-2008-313-C00042.

References

1 F. E. Browder, “Nonexpansive nonlinear operators in a Banach space,” Proceedings of the National Academy of Sciences of the United States of America, vol. 54, pp. 1041–1044, 1965.

2 W. A. Kirk, “A fixed point theorem for mappings which do not increase distances,” The American Mathematical Monthly, vol. 72, pp. 1004–1006, 1965.

3 L. Deng, “Convergence of the Ishikawa iteration process for nonexpansive mappings,” Journal of Mathematical Analysis and Applications, vol. 199, no. 3, pp. 769–775, 1996.

4 D. Lei and L. Shenghong, “Ishikawa iteration process with errors for nonexpansive mappings in uniformly convex Banach spaces,” International Journal of Mathematics and Mathematical Sciences, vol.

24, no. 1, pp. 49–53, 2000.

5 W. G. Dotson Jr., “On the Mann iterative process,” Transactions of the American Mathematical Society, vol. 149, pp. 65–73, 1970.

6 W. G. Dotson Jr., “Fixed points of quasi-nonexpansive mappings,” Journal of the Australian Mathematical Society, vol. 13, pp. 167–170, 1972.

7 S. Ishikawa, “Fixed points and iteration of a nonexpansive mapping in a Banach space,” Proceedings of the American Mathematical Society, vol. 59, no. 1, pp. 65–71, 1976.

8 W. V. Petryshyn and T. E. Williamson Jr., “Strong and weak convergence of the sequence of successive approximations for quasi-nonexpansive mappings,” Journal of Mathematical Analysis and Applications, vol. 43, pp. 459–497, 1973.

9 B. E. Rhoades, “Some fixed point iteration procedures,” International Journal of Mathematics and Mathematical Sciences, vol. 14, no. 1, pp. 1–16, 1991.

10 H. F. Senter and W. G. Dotson Jr., “Approximating fixed points of nonexpansive mappings,”

Proceedings of the American Mathematical Society, vol. 44, pp. 375–380, 1974.

11 K.-K. Tan and H. K. Xu, “Approximating fixed points of nonexpansive mappings by the Ishikawa iteration process,” Journal of Mathematical Analysis and Applications, vol. 178, no. 2, pp. 301–308, 1993.

12 T. Hu and J. Huang, “Convex hull of compact sets in a hyperspace,” Chinese Annals of Mathematics, vol. 20A, pp. 253–254, 1999Chinese.

13 Y. Xu, “Ishikawa and Mann iterative processes with errors for nonlinear strongly accretive operator equations,” Journal of Mathematical Analysis and Applications, vol. 224, no. 1, pp. 91–101, 1998.

14 S. B. Nadler Jr., “Multi-valued contraction mappings,” Pacific Journal of Mathematics, vol. 30, pp. 475–

488, 1969.

参照

関連したドキュメント

Zegeye, “On Mann and Ishikawa iteration schemes for multi-valued maps in Banach spaces,” Nonlinear Analysis: Theory, Methods &amp;

Ungchittrakool, “Strong convergence theorems of block iterative methods for a finite family of relatively nonexpansive mappings in Banach spaces,” Journal of Nonlinear and

Chang, “Viscosity approximation methods for a finite family of nonexpansive mappings in Banach spaces,” Journal of Mathematical Analysis and Applications, vol.. He,

Takahashi, “Strong convergence theorems by the hybrid method for families of mappings in Banach spaces,” Nonlinear Analysis: Theory, Methods &amp; Applications, vol..

Shahzad, Approximating fixed points of non-self nonexpansive mappings in Banach spaces, Nonlinear Analysis 61 (2005), no.. Chen, Viscosity approximation methods for

Takahashi, Weak and strong convergence theorems for nonexpansive mappings in Banach spaces, Nonlinear Anal.. Takahashi, A nonlinear ergodic theorem for an amenable semigroup

mixed equilibrium problem (1.1) and the set of common fixed points for an infinite family of $quasi-\phi$ -nonexpansive mappings in Banach

Soc., 74 (2006), 14&amp;151] for nonexpansive map- pings and establish strong convergence of such an iteration for asymptotically nonexpansive mappings in smooth and