Volume 2008, Article ID 484050,13pages doi:10.1155/2008/484050
Research Article
Composite Implicit General Iterative Process for
a Nonexpansive Semigroup in Hilbert Space
Lihua Li,1Suhong Li,1and Yongfu Su2
1Department of Mathematic and Physics, Hebei Normal University of Science and Technology
Qinhuangdao, Hebei 066004, China
2Department of Mathematics, Tianjin Polytechnic University, Tianjin 300160, China
Correspondence should be addressed to Lihua Li,[email protected]
Received 19 March 2008; Accepted 14 August 2008
Recommended by H´el`ene Frankowska
LetC be nonempty closed convex subset of real Hilbert spaceH. ConsiderC a nonexpansive semigroupI{Ts:s≥0}with a common fixed point, a contractionfwith coefficient 0< α <1, and a strongly positive linear bounded operatorAwith coefficientγ > 0. Let 0 < γ < γ/α. It is proved that the sequence{xn}generated iteratively byxn I−αnA1/tn
tn
0Tsynds
αnγfxn, yn I−βnAxnβnγfxnconverges strongly to a common fixed pointx∗ ∈ FI
which solves the variational inequalityγf−Ax∗, z−x∗ ≤0 for allz∈FI.
Copyrightq2008 Lihua Li 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
LetCbe a closed convex subset of a Hilbert spaceH, recall thatT :C→Cis nonexpansive if Tx−Ty ≤ x−y for allx, y ∈ C. Denote byFTthe set of fixed points ofT, that is,
FT:{x∈C:Txx}.
Recall that a familyI {Ts |0 ≤s <∞}of mappings fromCinto itself is called a nonexpansive semigroup onCif it satisfies the following conditions:
iT0xxfor allx∈C;
iiTst TsTtfor alls, t≥0;
iii Tsx−Tsy ≤ x−y for allx, y∈Cands≥0; ivfor allx∈C, s| →Tsxis continuous.
We denote by FI the set of all common fixed points of I, that is, FI
Iterative methods for nonexpansive mappings have recently been applied to solve convex minimization problemssee, e.g.,1–5and the references therein. A typical problem is to minimize a quadratic function over the set of the fixed points of a nonexpansive mapping on a real Hilbert spaceH:
min
x∈C
1
2Ax, x − x, b, 1.1
whereCis the fixed point set of a nonexpansive mappingT onH, andbis a given point in
H. Assume thatAis strongly positive, that is, there is a constantγ >0 with the property
Ax, x ≥γ x 2 ∀x∈H. 1.2
It is well known thatFTis closed convexcf.6. In3 see also4, it is proved that the sequence{xn}defined by the iterative method below, with the initial guessx0 ∈ Hchosen arbitrarily,
xn1
I−αnA
Txnαnb, n≥0 1.3
converges strongly to the unique solution of the minimization problem1.1provided that the sequence{αn}satisfies certain conditions.
On the other hand, Moudafi 7 introduced the viscosity approximation method for nonexpansive mappingssee8 for further developments in both Hilbert and Banach spaces. Let f be a contraction on H. Starting with an arbitrary initial x0 ∈ H, define a sequence{xn}recursively by
xn1
1−σn
Txnσnf
xn
, n≥0, 1.4
where{σn}is a sequence in0,1. It is proved7,8that under certain appropriate conditions
imposed on {σn}, the sequence{xn} generated by 1.4 strongly converges to the unique
solutionx∗inCof the variational inequality
I−fx∗, x−x∗≥0, x∈C. 1.5
Recently, Marino and Xu9combined the iterative method 1.3 with the viscosity approximation method1.4considering the following general iteration process:
xn1
I−αnA
Txnαnγf
xn
, n≥0, 1.6
generated by1.6converges strongly to the unique solution of the variational inequality
A−γfx∗, x−x∗≥0, x∈C, 1.7
which is the optimality condition for the minimization problem
min
x∈C
1
2Ax, x −hx, 1.8
wherehis a potential function forγfi.e.,hx γfx, forx∈H.
In this paper, motivated and inspired by the idea of Marino and Xu9, we introduce the composite implicit general iteration process1.9as follows:
xn
I−αnA
1
tn
tn
0
Tsyndsαnγf
xn
,
yn
I−βnA
xnβnγf
xn
,
1.9
where{αn},{βn} ⊂0,1, and investigate the problem of approximating common fixed point
of nonexpansive semigroup {Ts : s ≥ 0} which solves some variational inequality. The results presented in this paper extend and improve the main results in Marino and Xu9, and the methods of proof given in this paper are also quite different.
In what follows, we will make use of the following lemmas. Some of them are known; others are not hard to derive.
Lemma 1.1Marino and Xu9. Assume thatAis a strongly positive linear bounded operator on a Hilbert spaceHwith coefficientγ >0and0< ρ≤ A −1. Then I−ρA ≤1−γ.
Lemma 1.2Shimizu and Takashi10. LetCbe a nonempty bounded closed convex subset ofH and letI{Ts: 0≤s <∞}be a nonexpansive semigroup onC, then for anyh≥0,
lim
t→ ∞supx∈C
1tt
0
Tsxds−Th 1 t
t
0
Tsxds0. 1.10
Lemma 1.3. LetCbe a nonempty bounded closed convex subset of a Hilbert spaceHand let I
{Tt : 0 ≤ t < ∞} be a nonexpansive semigroup onC. If{xn}is a sequence inCsatisfying the following properties:
ixn z;
iilim supt→ ∞lim supn→ ∞ Ttxn−xn 0,
wherexn zdenote that{xn}converges weakly toz, thenz∈FI.
2. Main results
Lemma 2.1. LetHbe a Hilbert space,Ca closed convex subset ofH, letI {Ts : s ≥0}be a nonexpansive semigroup onC,{tn} ⊂0,∞is a sequence, thenI−1/tn
tn
0Tsdsis monotone.
Proof. In fact, for allx, y∈H,
x−y, I− 1 tn
tn
0
Tsds
x− I− 1
tn
tn
0
Tsds
y
x−y 2−
x−y, 1 tn
tn
0
Tsxds− 1 tn
tn
0
Tsyds
≥ x−y 2− x−y 1
tn
tn
0
Tsx−Tsyds
≥ x−y 2− x−y 2 0.
2.1
Theorem 2.2. Let Cbe nonempty closed convex subset of real Hilbert spaceH, suppose thatf :
C→Cis a fixed contractive mapping with coefficient 0 < α < 1, andI {Ts : s ≥ 0}is a nonexpansive semigroup onCsuch thatFIis nonempty, andAis a strongly positive linear bounded operator with coefficientγ >0,{αn},{βn} ⊂0,1,{tn} ⊂0,∞are real sequences such that
lim
n→ ∞αn0, βn◦
αn
, lim
n→ ∞tn∞, 2.2
then for any0< γ < γ/α, there is a unique{xn} ∈Csuch that
xn
I−αnA
1
tn
tn
0
Tsyndsαnγf
xn
,
yn
I−βnA
xnβnγf
xn
,
2.3
and the iteration process{xn}converges strongly to the unique solutionx∗∈FIof the variational inequalityγf−Ax∗, z−x∗ ≤0for allz∈FI.
Proof. Our proof is divided into five steps.
Sinceαn→0,βn→0 as n→ ∞, we may assume, with no loss of generality, thatαn < A −1,β
n< A −1for alln≥1.
i{xn}is bounded.
Firstly, we will show that the mappingTnf :C→Cdefined by
Tnf
I−αnA
1
tn
tn
0
TsI−βnA
βnγf
is a contraction. Indeed, fromLemma 1.1, we have for anyx, y∈Cthat
Tf
nx−Tnfy≤I−αnA1 tn
tn
0
Ts
I−βnA
xβnγfx
−TsI−βnA
yβnγfydsαnγfx−fy
≤1−αnγI−βnA
xβnγfx
−I−βnA
yβnγfyαnγα x−y
≤1−αnγI−βnA x−y βnγα x−y
αnγαx−y
≤1−αnγ
1−βn
γ−γα x−y αnγα x−y
1−αnγ
1−βn
γ−γααnγα
x−y
1−αn
γ−γα−1−αnγ
βn
γ−γα x−y
<1−αn
γ−γα x−y < x−y .
2.5
Letxn∈Cbe the unique fixed point ofTnf. Thus,
xn
I−αnA
1
tn
tn
0
Tsyndsαnγf
xn
,
yn
I−βnA
xnβnγf
xn
2.6
is well defined. Next, we will show that{xn}is bounded.
Pick anyz∈FIto obtain
xn−z
I−αnA
1
tn
tn
0
Tsynds−z
αn
γfxn
−Az
≤I−αnA
1
tn
tn
o
Tsyn−zdsαn
γfxn
−fzγfz−Az
≤1−αnγyn−zαn
γfxn
−fzγfz−Az,
2.7
xn−z≤
1−αnγyn−zαnγαxn−zαnγfz−Az. 2.8
Also
yn−z≤I−βnAxn−zβnγf
xn
−Az
≤1−βnγxn−zβnγαxn−zβnγfz−Az
1−βn
γ−γαxn−zβnγfz−Az.
Substituting2.9into2.8, we obtain that
xn−z≤ 1−αnγ
1−βn
γ−γαxn−z
1−αnγ
βnγfz−Az
αnγαxn−zαnγfz−Az
1−αnγ
1−βn
γ−γααnγαxn−z
1−αnγ
βnαnγfz−Az
1−γ−γααn
1−αnγ
βnxn−z
αn
1−αnγ
βnγfz−Az,
γ−γααn
1−αnγ
βnxn−z≤
αn
1−αnγ
βnγfz−Az,
xn−z≤ 1
γ−γαγfz−Az.
2.10
Thus{xn}is bounded.
iilimn→ ∞ xn−Tsxn 0.
Denote thatzn : 1/tn
tn
0Tsynds, since{xn}is bounded, zn−z ≤ yn−z and {Azn},{fxn}are also bounded, From2.6and limn→ ∞αn0, we have
xn−znαnγf
xn
−Azn−→0 n−→ ∞. 2.11
LetK{w∈C: w−z ≤1/γ−γα γfz−Az }, thenKis a nonempty bounded closed convex subset ofCandTs-invariant. Since{xn} ⊂KandKis bounded, there existsr >0
such thatK⊂Br, it follows fromLemma 1.2that
lim
n→ ∞zn−Tszn0 ∀s≥0. 2.12
From2.11and2.12, we have
lim
n→ ∞xn−Tsxn0. 2.13
iiiThere exists a subsequence{xnk}of{xn}such thatxnk x∗ ∈FIandx∗ is the
unique solution of the following variational inequality:
A−γfx∗, x∗−z≤0 ∀z∈FI. 2.14
Firstly since
yn−xnβnγf
xn
From conditionβn→0 and the boundedness of{xn}, we obtain that yn−xn →0. Again by
boundedness of{xn}, we know that there exists a subsequence{nk}of{n}such thatxnk x∗.
Thenynk x∗. FromLemma 1.3and stepii, we have thatx∗∈FI.
Next we will prove thatx∗solves the variational inequality2.14. Since
xn
I−αnA
1
tn
tn
0
Tsyndsαnγf
xn
, 2.16
we derive that
A−γfxn−
1
αn
I−αnA I−
1
tn
tn
0
Tsds
yn
1
αn
I−αnA
yn−
I−αnA
xn
.
2.17
It follows that, for allz∈FI,
A−γfxn, yn−z
− 1
αn
I−αnA I− 1 tn
tn
0
Tsds
yn, yn−z
1
αn
I−αnA
yn−
I−αnA
xn, yn−z
− 1
αn I−
1
tn
tn
0
Tsds
yn− I−
1
tn
tn
0
Tsds
z, yn−z
A I− 1
tn
tn
0
Tsds
yn, yn−z
1
αn
I−αnA
yn−
I−αnA
xn, yn−z
.
2.18
UsingLemma 2.1, we have from2.18that
A−γfxn, yn−z
≤
A I− 1
tn
tn
0
Tsds
yn, yn−z
1
αn
I−αnA
yn−xn
, yn−z
≤
A I− 1
tn
tn
0
Tsds
yn, yn−z
1
αnβn
γf
xn
−Axnyn−z.
2.19
Now replacingnin2.19withnkand lettingk→ ∞, we notice that
I− 1 tnk
t
nk
0
Tsds
and from conditionβn ◦αnand boundedness of{xn}, we have
1
αnk
βnkγf
xnk
−Axnkynk−z−→0. 2.21
Forx∗∈FI, we obtain
A−γfx∗, x∗−z≤0. 2.22
From 9, Theorem 3.2, we know that the solution of the variational inequality 2.14 is unique. That is,x∗∈FIis a unique solution of2.14.
iv
lim sup
n→ ∞
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
≤0, 2.23
wherex∗is obtained in stepiii.
To see this, there exists a subsequence{ni}of{n}such that
lim sup
n→ ∞
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
lim
i→ ∞
1
tni
tni
0
Tsynids−x∗, γf
x∗−Ax∗
,
2.24
we may also assume thatxni z, then1/tni
tni
0 Tsynids z, note from step ii that
z∈FIin virtue ofLemma 1.2. It follows from the variational inequality2.14that
lim sup
n→ ∞
1
tn
tn
0
Tsxnds−x∗, γf
x∗−Ax∗
z−x∗, γfx∗−Ax∗≤0. 2.25
So2.23holds thank to2.14.
vxn→x∗ n→ ∞.
Finally, we will provexn→x∗. Since
yn−x∗2
I−βnA
xn−x∗
βn
γfxn
−Ax∗2
≤I−βnAxn−x∗2βnγf
xn
−Ax∗2
≤1−βnγxn−x∗2βnγf
xn
−Ax∗2.
Next, we calculate
xn−x∗2
I−αnA
1
tn
tn
0
Tsynds−x∗
αn
γfxn
−Ax∗
2
I−αnA
1
tn
tn
0
Tsynds−x∗
2α2nγf
xn
−Ax∗2
2αn
I−αnA 1 tn
tn
0
Tsynds−x∗
, γfxn
−Ax∗
≤1−αnγ
2
yn−x∗2α2nγf
xn
−Ax∗2
2αn
I−αnA
1
tn
tn
0
Tsynds−x∗
, γfxn
−Ax∗.
2.27
Thus it follows from2.26that
xn−x∗2≤ 1−αnγ
2
1−βnγxn−x∗2
1−αnγ
2
βnγf
xn
−Ax∗2
α2nγf
xn
−Ax∗22αn
1
tn
tn
0
Tsynds−x∗, γf
xn
−Ax∗
−2α2n
A 1
tn
tn
0
Tsynds−x∗
, γfxn
−Ax∗
≤1−αnγ
2
1−βnγxn−x∗2
1−αnγ
2
βnγf
xn
−Ax∗2
α2nγf
xn
−Ax∗22αnγ
1
tn
tn
0
Tsynds−x∗, f
xn
−fx∗
2αn
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
−2α2
n A 1 tn tn 0
Tsynds−x∗
, γfxn
−Ax∗
≤1−αnγ
2
1−βnγxn−x∗22αnγαyn−x∗xn−x∗
1−αnγ
2
βnγf
xn
−Ax∗2
αn 2 1 tn tn 0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗
≤1−αnγ
2
1−βnγxn−x∗22αnγα
1−βn
γ−γαxn−x∗2
2αnγαβnγf
x∗−Ax∗xn−x∗
1−αnγ
2
βnγf
xn
−Ax∗2
αn 2 1 tn tn 0
Tsynds−x∗, γf
x∗−Ax∗
αnγf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗
1−αnγ
2
1−βnγxn−x∗22αnγα
1−βnγxn−x∗
2αnβnα2γ2xn−x∗22αnγαβnγf
x∗−Ax∗xn−x∗
1−αnγ
2
βnγf
xn
−Ax∗2
αn 2 1 tn tn 0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗
1−αnγ
2
2αnγα
1−βnγxn−x∗22αnβnα2γ2xn−x∗2
2αnγαβnγf
x∗−Ax∗xn−x∗
1−αnγ
2
βnγf
xn
−Ax∗2
αn 2 1 tn tn 0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗
<1−αnγ
22
αnγαxn−x∗22αnβnα2γ2xn−x∗2
2αnγαβnγf
x∗−Ax∗xn−x∗
1−αnγ
2
βnγf
xn
−Ax∗2
αn 2 1 tn tn 0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22αn
A 1
tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗
1−2γ−γααnxn−x∗22αnβnα2γ2xn−x∗2
2αnγαβnγf
x∗−Ax∗xn−x∗
1−αnγ
2
βnγf
xn
−Ax∗2
αn 2 1 tn tn 0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗γ2xn−x∗2
.
Thus
2γ−γαxn−x∗2≤2βnα2γ2xn−x∗22γαβnγf
x∗−Ax∗·xn−x∗
1−αnγ
2βn αn
γf
xn
−Ax∗2
2
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗γ2x
n−x∗2
≤βn
2α2γ2xn−x∗22γαγf
x∗−Ax∗·xn−x∗
βn αn
γf
xn
−Ax∗2
2
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
αn γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗γ2xn−x∗2
.
2.29
Since{xn}is bounded, we can take a constantL1, L2, L3>0 such that
L1 ≥2α2γ2xn−x∗22γαγf
x∗−Ax∗·xn−x∗,
L2 ≥γf
xn
−Ax∗2,
L3 ≥γf
xn
−Ax∗22A 1 tn
tn
0
Tsynds−x∗
·γfxn
−Ax∗γ2xn−x∗2
2.30
for alln≥0. It then follows from2.29that
2γ−γαxn−x∗2 ≤βnL1
βn αnL2
2
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
Then
2γ−γαlim sup
n→ ∞
xn−x∗2
≤lim sup
n→ ∞
βnL1
βn
αnL2αnL3
2 lim sup
n→ ∞
1
tn
tn
0
Tsynds−x∗, γf
x∗−Ax∗
.
2.32
From conditionαn→0,βn◦αnand2.23, we conclude that
lim sup
n→ ∞
xn−x∗2≤
0. 2.33
Soxn→x∗. This completes the proof of theTheorem 2.2.
It follows from the above proof thatTheorem 2.2is valid for nonexpansive mappings. Thus, we have that Corollaries2.3and2.4are two special cases ofTheorem 2.2.
Corollary 2.3. LetTbe a nonexpansive mapping from nonempty closed convex subsetCof a Hilbert spaceHtoC,{xn}is generated by the following algorithm:
xn
I−αnA
Tynαnγf
xn
,
yn
I−βnA
xnβnγf
xn
, 2.34
where{αn},{βn} ⊂0,1are real sequences such that
lim
n→ ∞αn0, βn◦
αn
, 2.35
then for any0 < γ < γ/α, the sequence{xn}above converges strongly to the unique solutionx∗ ∈ FTof the variational inequalityγf−Ax∗, z−x∗ ≤0for allz∈FT.
Corollary 2.4. LetTbe a nonexpansive mapping from nonempty closed convex subsetCof a Hilbert spaceHtoC,{xn}is generated by the following algorithm:
xn
I−αnA
Tynαnγf
xn
,
yn
I−βnA
xnβnγf
xn
,
2.36
where{αn}is a sequence in (0,1) satisfying the following condition:limn→ ∞αn0, then the sequence
{xn}converges strongly to the unique solutionx∗∈FTof the variational inequalityI−fx∗, x∗− z ≤0for allz∈FT.
References
2 H.-K. Xu, “Iterative algorithms for nonlinear operators,”Journal of the London Mathematical Society, vol. 66, no. 1, pp. 240–256, 2002.
3 H.-K. Xu, “An iterative approach to quadratic optimization,” Journal of Optimization Theory and Applications, vol. 116, no. 3, pp. 659–678, 2003.
4 I. Yamada, “The hybrid steepest descent method for the variational inequality problem over the intersection of fixed point sets of nonexpansive mappings,” in Inherently Parallel Algorithms in Feasibility and Optimization and Their Applications (Haifa, 2000), D. Butnariu, Y. Censor, and S. Reich, Eds., vol. 8 ofStudies in Computational Mathematics, pp. 473–504, North-Holland, Amsterdam, The Netherlands, 2001.
5 I. Yamada, N. Ogura, Y. Yamashita, and K. Sakaniwa, “Quadratic optimization of fixed points of nonexpansive mappings in Hilbert space,”Numerical Functional Analysis and Optimization, vol. 19, no. 1-2, pp. 165–190, 1998.
6 K. Goebel and W. A. Kirk,Topics in Metric Fixed Point Theory, vol. 28 ofCambridge Studies in Advanced Mathematics, Cambridge University Press, Cambridge, UK, 1990.
7 A. Moudafi, “Viscosity approximation methods for fixed-points problems,” Journal of Mathematical Analysis and Applications, vol. 241, no. 1, pp. 46–55, 2000.
8 H.-K. Xu, “Viscosity approximation methods for nonexpansive mappings,” Journal of Mathematical Analysis and Applications, vol. 298, no. 1, pp. 279–291, 2004.
9 G. Marino and H.-K. Xu, “A general iterative method fornonexpansive mappings in Hilbert spaces,” Journal of Mathematical Analysis and Applications, vol. 318, no. 1, pp. 43–52, 2006.
10 T. Shimizu and W. Takahashi, “Strong convergence to common fixed points of families of nonexpansive mappings,”Journal of Mathematical Analysis and Applications, vol. 211, no. 1, pp. 71–83, 1997.