• No results found

Strong Convergence Theorems of the CQ Method for Nonexpansive Semigroups

N/A
N/A
Protected

Academic year: 2020

Share "Strong Convergence Theorems of the CQ Method for Nonexpansive Semigroups"

Copied!
8
0
0

Loading.... (view fulltext now)

Full text

(1)

Volume 2007, Article ID 59735,8pages doi:10.1155/2007/59735

Research Article

Strong Convergence Theorems of the CQ Method for

Nonexpansive Semigroups

Huimin He and Rudong Chen

Received 25 January 2007; Accepted 19 March 2007

Recommended by Jerzy Jezierski

Motivated by T. Suzuki, we show strong convergence theorems of the CQ method for nonexpansive semigroups in Hilbert spaces by hybrid method in the mathematical programming. The results presented extend and improve the corresponding results of Kazuhide Nakajo and Wataru Takahashi (2003).

Copyright © 2007 H. He and R. Chen. 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

Throughout this paper, letHbe a real Hilbert space with inner product·,·and · . We usexnxto indicate that the sequence{xn}converges weakly tox. Similarly,xn→xwill symbolize strong convergence. we denote byNandR+the sets of nonnegative integers and nonnegative real numbers, respectively. letCbe a closed convex subset of a Hilbert spaceH, and LetT:C→Cbe a nonexpansive mapping (i.e.,Tx−T y ≤ x−yfor allx,y∈C). We use Fix(T) to denote the set of fixed points ofT; that is, Fix(T)= {x∈

C:x=Tx}. We know that Fix(T) is nonempty ifCis bounded, for more details see [1]. In [2], Shioji and Takahashi introduce in a Hilbert space the implicit iteration

xn=αnu+

1−αn 1

tn tn

0 T(s)xnds, n∈N, (1.1)

Where{αn}is a sequence in (0,1),{tn}is a sequence of positive real numbers divergent to , for eacht≥0 andu∈C. In 2003, Suzuki [3] is the first to introduce again in a Hilbert space the following implicit iteration process:

xn=αnu+

1−αn

Ttn

xn

(2)

for the nonexpansive semigroup case. In 2005, Xu [4] established a Banach space version of the sequence (1.2) of Suzuki [3], he proved that ifEis a uniformly convex Banach space with a weakly continuous duality map (e.g.,lpfor 1< p <), ifCis a closed convex sub-set ofE, and if{T(t) :t∈R+}is a nonexpansive semigroup on a closed convex subset Csuch that Fix(T)= ∅, then under certain appropriate assumptions made and the se-quencesαnandtnof the parameters, he showed that the sequencexnimplicitly defined by (1.2) for alln≥1 converges strongly to a member ofF=t≥0Fix(T(t)).

Recently, Chen and He [5] extend and improve the corresponding results of Suzuki [3], ifEis a reflexive Banach space which admits a weakly sequentially continuous duality mappingJfromEtoE∗, supposeCis a nonempty closed convex subset ofE. Let{T(t) : t∈R+}be a nonexpansive semigroup onCsuch thatF(T)= ∅, and f :C→Cis a fixed contraction onC. Let{αn}and{tn}be sequences of real numbers satisfying 0< αn<1, tn>0 and limn→∞tn=limn→∞αn/tn=0. Define a sequence{xn}inCby

xn=αnf

xn

+1−αn

Ttn

xn

, n≥1. (1.3)

Then{xn}converges strongly toq, asn→ ∞.qis the element ofF, such thatq is the unique solution inFto the following variational inequality:

(f −I)q,j(x−q)≤0 ∀x∈F(T). (1.4)

Some other results can be seen in [6–8].

Nakajo and Takahashi [9] introduced an iteration procedure for nonexpansive self-mappingsTonCas follows:

x0=x∈C,

yn=αnxn+1−αnTxn,

Cn=z∈C; yn−z xn−z , Qn=

z∈C;xn−z,x0−xn

0, xn+1=PCn∩Qn

x0

(1.5)

for eachn∈N∪ {0}, whereαn∈[0,a] for somea∈[0, 1), and{xn}converges strongly toPFix(T)x0

Let{T(t) :t∈R+}be a nonexpansive semigroup on a closed convex subsetC of a Hilbert spaceH, that is,

(1) for eacht∈R+,T(t) is a nonexpansive mapping onC; (2)T(0)x=xfor allx∈C;

(3)T(s+t)=T(s)◦T(t) for alls,t∈R+;

(4) for each x∈X, the mappingT(·)xfromR+intoCis continuous. We putF=

t≥0Fix(T(t)). We know thatFis nonempty ifCis bounded, see [10].

(3)

Nakajo and Takahashi [9] also introduced an iteration procedure for nonexpansive semi-group{T(t) :t∈R+}onCas follows:

x0=x∈C,

yn=αnxn+

1−αn 1

tn tn

0 T(s)xnds, Cn=

z∈C; yn−z xn−z ,

Qn=

z∈C;xn−z,x0−xn

0,

xn+1=PCn∩Qn

x0

(1.6)

for eachn∈N∪ {0}, whereαn∈[0,a] for some a∈[0, 1) and {tn} is a positive real number divergent sequence, and the sequence{xn}converges strongly toPFx0.

In 2006, Martinez-Yanes and Xu [11] employ Nakajo-Takahashi [9] idea and prove some strong convergence theorems for nonexpansive mappings and maximal monotone operators.

In this paper, we consider an iteration procedure for nonexpansive semigroups{T(t) : t∈R+}onCas follows:

x0=x∈C,

yn=αnxn+1−αnTtnxn,

Cn=z∈C; yn−z xn−z ,

Qn=

z∈C;xn−z,x0−xn

0,

xn+1=PCn∩Qn

x0

(1.7)

for eachn∈N∪ {0}, whereαn∈[0,a] for somea∈[0, 1) andtn≥0 limn→∞tn=0. then the sequence{xn}converges strongly toPFx0.

In the sequel, we will need the following definitions and results.

Definition 1.1. A Banach spaceEis said to satisfy Opial’s condition [12] if whenever{xn}

is a sequence inEwhich converges weakly tox, asn→ ∞, then

lim sup n→∞

xnx <lim sup n→∞

xny , yEwithx=y. (1.8)

It is well known that Hilbert space andlp(1< l <) space satisfy Opial’s condition [13].

Lemma1.2 [14]. LetC be a nonempty closed convex subset of a Hilbert spaceH. Given

x∈Handy∈C, theny=PCxif and only ifx−y,y−z ≥0,is satisfied for allz∈C.

Lemma1.3 [14,15]. Every Hilbert spaceHhas Radon-Riesz property or Kadets-Klee

(4)

2. Main results

Lemma2.1. LetCbe a closed convex subset of a Hilbert spaceH. Let{T(t) :t∈R+}be a

nonexpansive semigroup onCsuch thatF= ∅, and the sequence{xn}generated by (1.7),

whereαn∈[0,a]for somea∈[0, 1), Then{xn}is well defined andF⊂Cn∩Qnfor every

n∈N∪ {0}.

Proof. It is obvious thatCnis closed andQnis closed and convex for everyn∈N∪ {0}.

It follows from thatCnis convex for everyn∈N∪ {0}becauseyn−z ≤ xn−zis equivalent to

y n−xn

2

+ 2yn−xn,xn−z

0. (2.1)

So,Cn∩Qnis closed and convex for everyn∈N∪ {0}. Letu∈F. Then from y

n−u = αnxn+

1−αn

Ttn

xn−u ≤αn xn−u +

1−αn T

tn

xn−u xn−u .

(2.2)

we haveu∈Cnfor eachn∈N∪ {0}. So, we haveF⊂Cnfor alln∈N∪ {0}.

Next, we show by mathematical induction that{xn}is well defined andF⊂Cn∩Qn for everyn∈N∪ {0}. Forn=0, we havex0=x∈CandQ0=C, and henceF⊂C0∩Q0. Suppose thatxk is given and F⊂Ck∩Qk for some k∈N∪ {0}.There exists a unique elementxk+1∈Ck∩Qk such thatxk+1=PCk∩Qk(x0). Fromxk+1=PCk∩Qk(x0), it holds that

xk+1−z,x0−xk+1

0 (2.3)

for eachz∈Ck∩Qk. SinceF⊂Ck∩Qk, we getF⊂Qk+1, therefore we haveF⊂Ck+1 Qk+1.

The proof is completed.

Lemma2.2. LetCbe a closed convex subset of a Hilbert spaceH. Let{T(t) :t∈R+}be a

nonexpansive semigroup onCsuch thatF= ∅, and the sequence{xn}generated by (1.7),

whereαn∈[0,a]for somea∈[0, 1), Thenlimn→∞xn+1−xn =0.

Proof. At first, we show thatFis a closed convex subset ofC. SinceT(t) :C→C,t >0 is

nonexpansive, we claim thatFis closed. In fact, if pn⊂F=

t≥0Fix(T(t)),n≥1, such that limn→∞pn=p, then we have

T(t)p=lim

n→∞T(t)pn=nlim→∞pn=p ∀t∈R+. (2.4)

Thusp∈F.

Next, we show thatFis convex, we will use the following identity in Hilbert space:

tx+ (1t)y 2

(5)

which holds for allx,y∈Hand for allt∈[0, 1] indeed,

tx+ (1t)y 2

=t2x2+ (1t)2y2+ 2t(1t)x,y

=tx2+ (1t)y2+ 2t(1t)x,y

−t(1−t)x2t(1t)y2

=tx2+ (1t)y2t(1t)x2+y22x,y

=tx2+ (1t)y2t(1t)xy2.

(2.6)

Letp1,p2∈Fand for allt∈[0, 1],p=tp1+ (1−t)p2, then

p−p1=(1−t)p2−p1, p−p2=(1−t)p1−p2. (2.7)

From (2.5) and (2.7), we have

pT(t)p 2

= tp1−T(t)p

+ (1−t)p2−T(t)p 2

=t p1−T(t)p 2

+ (1−t) p2−T(t)p 2

−t(1−t) p1−p2 2

≤t p1−p 2+ (1−t) p2−p 2−t(1−t) p1−p2 2

=t(1−t)2 p1−p2 2+t2(1−t) p1−p2 2−t(1−t) p1−p2 2

=t(1−t)(1−t+t−1) p1−p2 2

=0.

(2.8)

Thusp=T(t)p, for allt >0, that is,p∈F.

Secondly, we show that{xn}is bounded. SinceFis a nonempty closed convex subset ofC, there exists a unique elementz0∈Fsuch thatz0=PF(x0).Fromxn+1=PCn∩Qn(x0), we have

x

n+1−x0 z−x0 ∀z∈Cn∩Qn. (2.9)

It follows fromLemma 2.1thatF⊂Cn∩Qnfor everyn∈N∪ {0}, together withz0 F(T), we have

xn+1x0 z0x0 nN∪ {0}. (2.10)

This implies that{xn}is bounded, so{T(t)xn}is also bounded, and moreover so is {yn}sinceyn ≤αnxn+ (1−αn)T(t)xn.

Thirdly, we show thatxn+1−xn →0 asn→ ∞. SinceQn= {z∈C;xn−z,x0−xn ≥ 0},xn=PQn(x0). Asxn+1∈Cn∩Qn⊂Qn, we obtain

x

(6)

Therefore the sequence{xn−x0}is bounded and nondecreasing. So

lim

n→∞ xn−x0 exists. (2.12)

On the other hand, fromxn+1∈Qn, we getxn−xn+1,x0−xn ≥0, and hence

xnxn+1 2

= xn−x0−xn+1−x0 2

= xn−x0 2

2xn−x0,xn+1−x0

+ xn+1−x0 2

= xn−x0 2

+ xn+1−x0 2

2xn−x0,xn+1−xn+xn−x0

= xn+1−x0 2

xn−x0 2

2xn−xn+1,x0−xn

xn+1−x0 2

xn−x0 2

−→0 (n−→ ∞).

(2.13)

So

lim

n→∞ xn+1−xn =0. (2.14)

This proof is completed.

Theorem2.3. LetCbe a closed convex subset of a Hilbert space H. Let{T(t) :t∈R+}

be a nonexpansive semigroup onCsuch thatF= ∅, and the sequence{xn}generated by

(1.7), whereαn∈[0,a]for some a∈[0, 1), andtn≥0 limn→∞tn=0. then the sequence

{xn}converges strongly toPFx0.

Proof. It follows fromxn+1∈Cnthat

T

tn

xn−xn =11 αn

ynxn

1

1−αn

ynxn+1 + x

n+1−xn

2

1−αn

xn+1xn

(2.15)

for everyn∈N∪ {0}. ByLemma 2.2, we getT(tn)xn−xn →0.

We claim that{xn} is relatively sequentially compact. Indeed, there exists a weakly convergence subsequence{xnj} ⊆ {xn} by reflexivity ofH and boundedness of the se-quence{xn}, now we supposexnjx∈C(j→ ∞). Now we show thatx∈F. Putxj=xnj, βj=αnj, andsj=tnj forj∈N, letsj≥0 be such that

sj−→0, T

sjxj−xj

(7)

Fixt >0, from

x

j−T(t)x

[t/sj]−1

k=0

T

(k+ 1)sj

xj−T

ksj

xj

+ T

t sj sj

xj−T t sj sj

x + Tt/sj

sj

x−T(t)x

t sj T

sjxj−xj + xj−x + T t− t sj sj

x−x

≤t T

sj

xj−xj

sj +

x

j−x + max T(s)x−x : 0≤s≤sj

.

(2.17)

For allj∈N∪ {0}, as every Hilbert space satisfies Opial’s condition, then we have

lim sup j→∞

xjT(t)x lim sup j→∞

xjx .

(2.18)

This implies thatT(t)x=x. Therefore,

x∈F. (2.19)

Ifz0=PF(x0), it follows from (2.10), (2.19), and the lower semicontinuity of the norm that

x0z0 x0x lim inf

j→∞ x0−xnj lim sup j→∞

x0xn

j x0−z0 . (2.20)

Thus, we obtain

lim

j→∞ xnj−x0 = x0−x = x0−z0 . (2.21)

This implies that

xnj−→x=z0. (2.22)

This shows that{xn}is relatively sequentially compact. Therefore, we havexn→z0.

The proof is completed.

Acknowledgment

(8)

References

[1] F. E. Browder, “Fixed-point theorems for noncompact mappings in Hilbert space,”Proceedings of the National Academy of Sciences of the United States of America, vol. 53, no. 6, pp. 1272–1276, 1965.

[2] N. Shioji and W. Takahashi, “Strong convergence theorems for asymptotically nonexpansive semigroups in Hilbert spaces,”Nonlinear Analysis, vol. 34, no. 1, pp. 87–99, 1998.

[3] T. Suzuki, “On strong convergence to common fixed points of nonexpansive semigroups in Hilbert spaces,”Proceedings of the American Mathematical Society, vol. 131, no. 7, pp. 2133–2136, 2003.

[4] H.-K. Xu, “A strong convergence theorem for contraction semigroups in Banach spaces,”Bulletin of the Australian Mathematical Society, vol. 72, no. 3, pp. 371–379, 2005.

[5] R. Chen and H. He, “Viscosity approximation of common fixed points of nonexpansive semi-groups in Banach space,” to appear inApplied Mathematics Letters.

[6] R. Chen and Z. Zhu, “Viscosity approximation fixed points for nonexpansive andm-accretive operators,”Fixed Point Theory and Applications, vol. 2006, Article ID 81325, 10 pages, 2006. [7] R. Chen, Y. Song, and H. Zhou, “Convergence theorems for implicit iteration process for a

fi-nite family of continuous pseudocontractive mappings,”Journal of Mathematical Analysis and Applications, vol. 314, no. 2, pp. 701–709, 2006.

[8] Y. Yao and R. Chen, “Iterative algorithm for approximating solutions of maximal monotone operators in Hilbert spaces,”Fixed Point Theory and Applications, vol. 2007, Article ID 32870, 8 pages, 2007.

[9] K. Nakajo and W. Takahashi, “Strong convergence theorems for nonexpansive mappings and nonexpansive semigroups,”Journal of Mathematical Analysis and Applications, vol. 279, no. 2, pp. 372–379, 2003.

[10] 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, no. 4, pp. 1041–1044, 1965. [11] C. Martinez-Yanes and H.-K. Xu, “Strong convergence of the CQ method for fixed point

itera-tion processes,”Nonlinear Analysis, vol. 64, no. 11, pp. 2400–2411, 2006.

[12] Z. Opial, “Weak convergence of the sequence of successive approximations for nonexpansive mappings,”Bulletin of the American Mathematical Society, vol. 73, pp. 591–597, 1967.

[13] K. Yanagi, “On some fixed point theorems for multivalued mappings,”Pacific Journal of Mathe-matics, vol. 87, no. 1, pp. 233–240, 1980.

[14] R. E. Megginson,An Introduction to Banach Space Theory, vol. 183 ofGraduate Texts in Mathe-matics, Springer, New York, NY, USA, 1998.

[15] W. Takahashi,Nonlinear Functional Analysis. Fixed Point Theory and Its Applications, Yokohama Publishers, Yokohama, Japan, 2000.

Huimin He: Department of Mathematics, Tianjin Polytechnic University, Tianjin 300160, China Email address:[email protected]

References

Related documents

While many studies have demonstrated that viruses re- program cell metabolism and rely on metabolic changes for optimal virus replication in vitro, significant work re- mains

The effects of excessive concentration are: overaccumulation of capital in the national economy; financialisation of the national economy (the transformation into the

Western blot and immunofluorescence experiments confirm that con- jugated SUMO 2 is present at over 10 times higher levels in FXTAS patient brain nuclei compared with control

Herein, we examine the responses of two key molecular markers of dietary stress, namely variant-specific methylation at ribosomal DNA (rDNA) and small RNA distribution, including

Thus, this paper presents the result of an experimental investigation on predicting surface roughness (Ra) of tool steel using Taguchi’s parameter design and response

To determine if the most common and burdensome muscu- loskeletal conditions (neck or back pain or OA of the knee or hip) increased the development of chronic dis- ease

Mean intensity of mEGFP- α -tubulin signal and cell diameter was measured in cells before and after control ( blue , seven cells), hypo- ( red , eight cells) or hyper- ( green ,

If one or more qualifying surveys were available for a GBD sub-region, the data from those were utilised to arrive at the best estimates for the proportion of blindness (pre-