Open Access

Algorithmic and analytical approach to the split common fixed points problem

Fixed Point Theory and Applications20142014:172

https://doi.org/10.1186/1687-1812-2014-172

Received: 14 March 2014

Accepted: 25 July 2014

Published: 18 August 2014

Abstract

The split problem, especially the split common fixed point problem, has been studied by many authors. In this paper, we study the split common fixed point problem for the pseudo-contractive mappings and the quasi-nonexpansive mappings. We suggest and analyze an iterative algorithm for solving this split common fixed point problem. A weak convergence theorem is given.

MSC:49J53, 49M37, 65K10, 90C25.

Keywords

split common fixed point problem pseudo-contractive mappings quasi-nonexpansive mapping weak convergence

1 Introduction

This article we devote to the split common fixed point problem and study it for the pseudo-contractive and quasi-nonexpansive mappings. The split common fixed point problem is a generalization of the convex feasibility problem which is to find a point x satisfying the following:
x i = 1 m C i ,
where m 1 is an integer, and each C i is a nonempty closed convex subset of a Hilbert space H. Note that the convex feasibility problem has received a lot of attention due to its extensive applications in many applied disciplines as diverse as approximation theory, image recovery and signal processing, control theory, biomedical engineering, communications, and geophysics (see [13] and the references therein). A special case of the convex feasibility problem is the split feasibility problem, which is to find a point x such that
x C and A x Q ,
(1.1)
where C and Q are two closed convex subsets of two Hilbert spaces H 1 and H 2 , respectively, and A : H 1 H 2 is a bounded linear operator. Such problems arise in the field of intensity-modulated radiation therapy when one attempts to describe physical dose constraints and equivalent uniform dose constraints within a single model; see [4]. The problem with only a single pair of sets C R N and Q R M was first introduced by Censor and Elfving [5]. They used their simultaneous multi-projections algorithm to solve the split feasibility problem. Their algorithms, as well as others, see, e.g., Byrne [6], involve matrix inversion at each iterative step. Calculating inverses of matrices is very time-consuming, particularly if the dimensions are large. Therefore, a new algorithm for solving the split feasibility problem was devised by Byrne [7], called the CQ-algorithm:
x n + 1 = P C ( x n τ A ( I P Q ) A x n ) ,

where τ ( 0 , 2 L ) with L being the largest eigenvalue of the matrix A A , I is the unit matrix or operator and P C and P Q denote the orthogonal projections onto C and Q, respectively. In the case of nonlinear constraint sets, orthogonal projections may demand a great amount of work of solving a nonlinear optimization problem to minimize the distance between the point and the constraint set. However, it can easily be estimated by linear approximation using the current constraint violation and the sub-gradient at the current point. This was done by Yang, in his recent paper [8], where he proposed a relaxed version of the CQ-algorithm in which orthogonal projections are replaced by sub-gradient projections, which are easily executed when the sets C and Q are given as lower level sets of convex functions; see also [9]. There are a large number of references on the CQ method for the split feasibility problem in the literature; see, for instance, [1025].

It is our main purpose in this paper to develop algorithms for the split common fixed point for the pseudo-contractive and quasi-nonexpansive mappings. Weak convergence theorem is given. Our results improve and develop previously discussed feasibility problems and related algorithms.

2 Preliminaries

Let H be a real Hilbert space with inner product , and norm , respectively. Let C be a nonempty closed convex subset of H.

Definition 2.1 A mapping S : C C is called pseudo-contractive if
S x S y , x y x y 2

for all x , y C .

We will use Fix ( S ) to denote the set of fixed points of S, that is, Fix ( S ) = { x C : x = S x } .

Remark 2.2 It is clear that S is pseudo-contractive if and only if
S x S y 2 x y 2 + ( I S ) x ( I S ) y 2
(2.1)

for all x , y C .

Interest in pseudo-contractive mappings stems mainly from their firm connection with the class of nonlinear accretive operators. It is a classical result, see Deimling [26], that if S is an accretive operator, then the solutions of the equations S x = 0 correspond to the equilibrium points of some evolution systems.

Definition 2.3 A mapping T : C C is called L-Lipschitzian if there exists L > 0 such that
T x T y L x y

for all x , y C .

Remark 2.4 We call T nonexpansive if L = 1 and T is contractive if L < 1 .

Definition 2.5 A mapping T : C C is called quasi-nonexpansive if
T x x x x , ( x , x ) C × Fix ( T ) .

Remark 2.6 It is obvious that if T is nonexpansive with Fix ( T ) , then T is quasi-nonexpansive.

Usually, the convergence of fixed point algorithms requires some additional smoothness properties of the mapping T such as demi-closedness.

Definition 2.7 A mapping T is said to be demi-closed if, for any sequence { x k } which weakly converges to x ˜ , and if the sequence { T ( x k ) } strongly converges to z, then T ( x ˜ ) = z .

It is well known that in a real Hilbert space H, the following equality holds:
t x + ( 1 t ) y 2 = t x 2 + ( 1 t ) y 2 t ( 1 t ) x y 2
(2.2)

for all x , y H and t [ 0 , 1 ] .

Lemma 2.8 ([27])

Let H be a real Hilbert space, C a closed convex subset of H. Let U : C C be a continuous pseudo-contractive mapping. Then
  1. (i)

    Fix ( U ) is a closed convex subset of C,

     
  2. (ii)

    ( I U ) is demi-closed at zero.

     

Lemma 2.9 ([28])

Let H be a Hilbert space and let { u n } be a sequence in H such that there exists a nonempty set Ω H satisfying the following:
  1. (i)

    for every u Ω , lim n u n u exists,

     
  2. (ii)

    any weak-cluster point of the sequence { u n } belongs in Ω.

     

Then there exists x Ω such that { u n } weakly converges to x .

In the sequel we shall use the following notation:
  1. 1.

    ω w ( u n ) = { x : u n j x  weakly } denote the weak ω-limit set of { u n } ;

     
  2. 2.

    u n x stands for the weak convergence of { u n } to x;

     
  3. 3.

    u n x stands for the strong convergence of { u n } to x.

     

3 Main results

In this section, we will focus our attention on the following general two-operator split common fixed point problem:
find  x C  such that  A x Q ,
(3.1)
where A : H 1 H 2 is a bounded linear operator, U : H 1 H 1 is a pseudo-contractive mapping and T : H 2 H 2 is a quasi-nonexpansive mapping with nonempty fixed point sets Fix ( U ) = C and Fix ( T ) = Q , and we denote the solution set of the two-operator split common fixed point problem by
Γ = { x C ; A x Q } .
To solve (3.1), Censor and Segal [12] proposed and proved, in finite-dimensional spaces, the convergence of the following algorithm:
x k + 1 = U ( x k + γ A ( T I ) A x k ) , k N ,
(3.2)

where γ ( 0 , 2 λ ) , with λ being the largest eigenvalue of the matrix A A .

Moudafi [16] extended (3.2) to the following relaxed algorithm:
x k + 1 = U α k ( x k + γ A ( T β I ) A x k ) , k N ,

where β ( 0 , 1 ) , α k ( 0 , 1 ) are relaxation parameters.

Inspired by their works, we introduce the following algorithm.

Algorithm 3.1 Let H 1 and H 2 be two real Hilbert spaces. Let A : H 1 H 2 be a bounded linear operator. Let U : H 1 H 1 be a pseudo-contractive mapping with Lipschitzian constant L and T : H 2 H 2 be a quasi-nonexpansive mapping with nonempty Fix ( U ) = C and Fix ( T ) = Q . Let x 0 H 1 . Define a sequence { u n } as follows:
{ x n = u n + γ ν A ( T I ) A u n , y n = ( 1 ξ n ) x n + ξ n U x n , u n + 1 = [ 1 ( 1 δ n ) α n ] x n + ( 1 δ n ) α n U y n
(3.3)

for all n N , where γ and ν are two constants, { α n } , { δ n } , and { ξ n } are three sequences in [ 0 , 1 ] .

In the sequel, we assume the parameters satisfy the following restrictions.

Parameters restrictions:

(R1): 0 < ν < 1 and 0 < γ < 1 λ ν , where λ is the largest eigenvalue of the matrix A A ;

(R2): 0 < lim inf n α n lim sup n α n < 1 ;

(R3): 0 < k 1 δ n ξ n < 1 1 + L 2 + 1 for all n N , where L is the Lipschitz constant of U.

Remark 3.2 Without loss of generality, we may assume that the Lipschitz constant L > 1 . It is obvious that 1 1 + L 2 + 1 < 1 L for all n 1 . Since ξ n < 1 1 + L 2 + 1 , we have 1 2 ξ n ξ n 2 L 2 > 0 for all n N .

Theorem 3.3 Let H 1 and H 2 be two real Hilbert spaces. Let A : H 1 H 2 be a bounded linear operator. Let U : H 1 H 1 be a pseudo-contractive mapping with Lipschitzian constant L and T : H 2 H 2 be a quasi-nonexpansive mapping with nonempty Fix ( U ) = C and Fix ( T ) = Q . Assume that T I is demi-closed at 0 and Γ . Then the sequence { u n } generated by algorithm (3.3) weakly converges to a split common fixed point μ Γ .

Proof Let x Γ . Then we get x Fix ( U ) and A x Fix ( T ) . From (2.2) and (3.3), we have
u n + 1 x 2 = [ 1 ( 1 δ n ) α n ] x n + ( 1 δ n ) α n U y n x 2 = ( 1 α n ) ( x n x ) + α n [ δ n x n + ( 1 δ n ) U y n x ] 2 = ( 1 α n ) x n x 2 + α n δ n x n + ( 1 δ n ) U y n x 2 α n ( 1 α n ) δ n x n + ( 1 δ n ) U y n x n 2 = α n [ δ n x n x 2 + ( 1 δ n ) U y n x 2 δ n ( 1 δ n ) U y n x n 2 ] + ( 1 α n ) x n x 2 α n ( 1 α n ) δ n x n + ( 1 δ n ) U y n x n 2 .
(3.4)
Since x Fix ( U ) , we have from (2.1)
U x x 2 x x 2 + x U x 2
(3.5)

for all x C .

By (3.4) and (3.5), we obtain
U y n x 2 = U ( ( 1 ξ n ) x n + ξ n U x n ) x 2 ( 1 ξ n ) x n + ξ n U x n U y n 2 + ( 1 ξ n ) x n + ξ n U x n x 2 = ( 1 ξ n ) ( x n U y n ) + ξ n ( U x n U y n ) 2 + ( 1 ξ n ) ( x n x ) + ξ n ( U x n x ) 2 = ( 1 ξ n ) x n U y n 2 + ξ n U x n U y n 2 ξ n ( 1 ξ n ) x n U x n 2 + ( 1 ξ n ) x n x 2 + ξ n U x n x 2 ξ n ( 1 ξ n ) x n U x n 2 ( 1 ξ n ) x n x 2 + ξ n ( x n x 2 + x n U x n 2 ) ξ n ( 1 ξ n ) x n U x n 2 + ( 1 ξ n ) x n U y n 2 + ξ n U x n U y n 2 ξ n ( 1 ξ n ) x n U x n 2 .
Note that U is L-Lipschitzian and x n y n = ξ n ( U x n x n ) . Then we have
U y n x 2 ( 1 ξ n ) x n x 2 + ξ n ( x n x 2 + x n U x n 2 ) ξ n ( 1 ξ n ) x n U x n 2 + ( 1 ξ n ) x n U y n 2 + ξ n 3 L 2 x n U x n 2 ξ n ( 1 ξ n ) x n U x n 2 = x n x 2 + ( 1 ξ n ) x n U y n 2 ξ n ( 1 2 ξ n ξ n 2 L 2 ) x n U x n 2 .
(3.6)
Substituting (3.6) into (3.4), we have
u n + 1 x 2 ( 1 α n ) x n x 2 + α n { δ n x n x 2 + ( 1 δ n ) [ x n x 2 + ( 1 ξ n ) x n U y n 2 ξ n ( 1 2 ξ n ξ n 2 L 2 ) x n U x n 2 ] δ n ( 1 δ n ) U y n x n 2 } α n ( 1 α n ) δ n x n + ( 1 δ n ) U y n x n 2 x n x 2 α n ( 1 α n ) δ n x n + ( 1 δ n ) U y n x n 2 .
(3.7)
Since λ is the spectral radius of the operator A A , we deduce
( T I ) A u n , A A ( T I ) A u n λ ( T I ) A u n 2 .
This together with (3.1) implies that
x n x 2 = u n + γ ν A ( T I ) A u n x 2 = u n x 2 + 2 γ ν A ( T I ) A u n , u n x + γ 2 ν 2 A ( T I ) A u n 2 = u n x 2 + 2 γ ν A ( T I ) A u n , u n x + γ 2 ν 2 ( T I ) A u n , A A ( T I ) A u n u n x 2 + 2 γ ν A ( T I ) A u n , u n x + γ 2 ν 2 λ ( T I ) A u n 2 .
(3.8)
Since T is quasi-nonexpansive and A x Fix ( T ) , we have
T A u n A x A u n A x .
At the same time, we have the following equality in Hilbert spaces:
x y 2 = x 2 + y 2 2 x , y .
(3.9)
In (3.9), picking up x = ( T I ) A u n and y = T A u n A x to deduce
A u n A x 2 = ( T I ) A u n ( T A u n A x ) 2 = ( T I ) A u n 2 + T A u n A x 2 2 ( T I ) A u n , T A u n A x ( T I ) A u n 2 + A u n A x 2 2 ( T I ) A u n , T A u n A x .
It follows that
( T I ) A u n , T A u n A x 1 2 ( T I ) A u n 2 .
Thus,
A ( T I ) A u n , u n x = ( T I ) A u n , A u n A x = ( T I ) A u n , T A u n A x + ( T I ) A u n , A u n T A u n 1 2 ( T I ) A u n 2 ( T I ) A u n 2 = 1 2 ( T I ) A u n 2 .
(3.10)
From (3.7), (3.8), and (3.10), we get
u n + 1 x 2 u n x 2 γ ν ( 1 λ γ ν ) ( T I ) A u n 2 α n ( 1 α n ) δ n x n + ( 1 δ n ) U y n x n 2 .
(3.11)
We deduce immediately that
u n + 1 x u n x .
Hence, lim n u n x exists. This implies that { u n } is bounded. Consequently, we have
0 γ ν ( 1 λ γ ν ) ( T I ) A u n 2 u n x 2 u n + 1 x 2 0 .
Therefore,
lim n ( T I ) A u n = 0 .
(3.12)
Since { u n } is bounded, ω w ( u n ) . We can take μ ω w ( u n ) , that is, there exists { u n j } such that ω lim j u n j = μ . Noting that T I is demi-closed at 0, from (3.12), we obtain
( T I ) A μ = 0 .

Thus, A μ Fix ( T ) .

From (3.11), we deduce
α n ( 1 α n ) δ n x n + ( 1 δ n ) U y n x n 2 u n x 2 u n + 1 x 2 0 .
Since 0 < lim inf n α n lim sup n α n < 1 , 0 < lim inf n α n lim sup n α n ( 1 α n ) < 1 . Then we have
lim n δ n x n + ( 1 δ n ) U y n x n = lim n ( 1 δ n ) U y n x n = 0 .
Note that lim sup δ n < 1 ; we get immediately
lim n U y n x n = 0 .
Since U is L-Lipschitzian, we have
U x n x n U x n U y n + U y n x n L x n y n + U y n x n = L ξ n U x n x n + U y n x n .
It follows that
U x n x n 1 1 L ξ n U y n x n 1 + L 2 + 1 L 1 + L 2 + 1 U y n x n .
So
lim n U x n x n = 0 .
(3.13)
From (3.3), (3.12), and (3.13), we have lim n x n u n = 0 . Thus, ω lim j x n j = μ . By the demi-closedness of U I at 0 (Lemma 2.8), we get
U μ = μ .

Hence, μ Fix ( U ) . Therefore, μ Γ . Since there is no more than one weak-cluster point, the weak convergence of the whole sequence { u n } follows by applying Lemma 2.9 with Ω = Γ . This completes the proof. □

Example 3.4 Let H = R with the inner product defined by x , y = x y for all x , y R and the standard norm | | . Let C = [ 0 , + ) and T x = x 2 + 2 1 + x for all x C . Obviously, Fix ( T ) = 2 . It is easy to see that
| T x 2 | = | x 2 + 2 1 + x 2 | = x 1 + x | x 2 | | x 2 |
for all x C and
| T ( 0 ) T ( 1 3 ) | = 5 12 > | 0 1 3 | .

Hence, T is a continuous quasi-nonexpansive mapping but not nonexpansive.

Example 3.5 Let H = R with the inner product defined by x , y = x y for all x , y R and the standard norm | | . Let C = [ 0 , + ) and let U x = x 1 + 4 x + 1 for all x C . Observe that Fix ( U ) = 3 . It is easy to see that
U x U y , x y = x 1 + 4 x + 1 y + 1 4 y + 1 , x y [ 1 4 ( x + 1 ) ( y + 1 ) ] | x y | 2 | x y | 2
and
| U x U y | | x 1 + 4 x + 1 y + 1 4 y + 1 | | 1 4 ( x + 1 ) ( y + 1 ) | | x y | 5 | x y |

for all x , y C .

But
| U ( 1 4 ) U ( 0 ) | = 11 20 > 1 4 .

Hence, U is a Lipschitzian pseudo-contractive mapping but it is not nonexpansive.

Declarations

Acknowledgements

The authors are grateful to the five reviewers for their valuable comments and suggestions. Li-Jun Zhu was supported in part by NNSF of China (61362033) and NZ13087. Yeong-Cheng Liou was supported in part by NSC 101-2628-E-230-001-MY3 and NSC 101-2622-E-230-005-CC3.

Authors’ Affiliations

(1)
School of Mathematics and Information Science, Beifang University of Nationalities
(2)
Department of Information Management, Cheng Shiu University
(3)
Center for General Education, Kaohsiung Medical University
(4)
Department of Mathematics and RINS, Gyeongsang National University
(5)
Department of Mathematics, Tianjin Polytechnic University

References

  1. Bauschke HH, Borwein JM: On projection algorithms for solving convex feasibility problems. SIAM Rev. 1996, 38: 367–426. 10.1137/S0036144593251710View ArticleMathSciNetGoogle Scholar
  2. Combettes PL: The convex feasibility problem in image recovery. 95. In Advances in Imaging and Electron Physics. Edited by: Hawkes P. Academic Press, New York; 1996:155–270.Google Scholar
  3. Stark H (Ed): Image Recovery Theory and Applications. Academic Press, Orlando; 1987.Google Scholar
  4. Censor Y, Bortfeld T, Martin B, Trofimov A: A unified approach for inversion problems in intensity modulated radiation therapy. Phys. Med. Biol. 2006, 51: 2353–2365. 10.1088/0031-9155/51/10/001View ArticleGoogle Scholar
  5. Censor Y, Elfving T: A multiprojection algorithm using Bregman projections in a product space. Numer. Algorithms 1994, 8: 221–239. 10.1007/BF02142692View ArticleMathSciNetGoogle Scholar
  6. Byrne C: Bregman-Legendre multidistance projection algorithms for convex feasibility and optimization. In Inherently Parallel Algorithms in Feasibility and Optimization and Their Applications. Edited by: Butnariu D, Censor Y, Reich S. Elsevier, Amsterdam; 2001:87–100.View ArticleGoogle Scholar
  7. Byrne C: Iterative oblique projection onto convex subsets and the split feasibility problem. Inverse Probl. 2002, 18: 441–453. 10.1088/0266-5611/18/2/310View ArticleMathSciNetGoogle Scholar
  8. Yang Q: The relaxed CQ algorithm solving the split feasibility problem. Inverse Probl. 2004, 20: 1261–1266. 10.1088/0266-5611/20/4/014View ArticleGoogle Scholar
  9. Zhao J, Yang Q: Several solution methods for the split feasibility problem. Inverse Probl. 2005, 21: 1791–1799. 10.1088/0266-5611/21/5/017View ArticleGoogle Scholar
  10. Ceng LC, Ansari QH, Yao JC: An extragradient method for split feasibility and fixed point problems. Comput. Math. Appl. 2012, 64: 633–642. 10.1016/j.camwa.2011.12.074View ArticleMathSciNetGoogle Scholar
  11. Censor Y, Motova A, Segal A: Perturbed projections and subgradient projections for the multiple-sets split feasibility problem. J. Math. Anal. Appl. 2007, 327: 1244–1256. 10.1016/j.jmaa.2006.05.010View ArticleMathSciNetGoogle Scholar
  12. Censor Y, Segal A: The split common fixed point problem for directed operators. J. Convex Anal. 2009, 16: 587–600.MathSciNetGoogle Scholar
  13. Dang Y, Gao Y: The strong convergence of a KM-CQ-like algorithm for a split feasibility problem. Inverse Probl. 2011., 27: Article ID 015007 10.1088/0266-5611/27/1/015007Google Scholar
  14. Deepho J, Kumam P: A modified Halpern’s iterative scheme for solving split feasibility problems. Abstr. Appl. Anal. 2012., 2012: Article ID 876069 10.1155/2012/876069Google Scholar
  15. López G, Martín-Márquez V, Xu HK: Iterative algorithms for the multiple-sets split feasibility problem. In Biomedical Mathematics: Promising Directions in Imaging, Therapy Planning and Inverse Problems. Edited by: Censor Y, Jiang M, Wang G. Medical Physics Publishing, Madison; 2009:243–279.Google Scholar
  16. Moudafi A: A note on the split common fixed-point problem for quasi-nonexpansive operators. Nonlinear Anal. 2011, 74: 4083–4087. 10.1016/j.na.2011.03.041View ArticleMathSciNetGoogle Scholar
  17. Wang F, Xu HK: Approximating curve and strong convergence of the CQ algorithm for the split feasibility problem. J. Inequal. Appl. 2010., 2010: Article ID 102085 10.1155/2010/102085Google Scholar
  18. Wang F, Xu HK: Cyclic algorithms for split feasibility problems in Hilbert spaces. Nonlinear Anal. 2011, 74: 4105–4111. 10.1016/j.na.2011.03.044View ArticleMathSciNetGoogle Scholar
  19. Xu HK: A variable Krasnosel’skii-Mann algorithm and the multiple-set split feasibility problem. Inverse Probl. 2006, 22: 2021–2034. 10.1088/0266-5611/22/6/007View ArticleGoogle Scholar
  20. Xu HK: Iterative methods for the split feasibility problem in infinite-dimensional Hilbert spaces. Inverse Probl. 2010., 26: Article ID 105018 10.1088/0266-5611/26/10/105018Google Scholar
  21. Yao Y, Kim TH, Chebbi S, Xu HK: A modified extragradient method for the split feasibility and fixed point problems. J. Nonlinear Convex Anal. 2012, 13: 383–396.MathSciNetGoogle Scholar
  22. Yao Y, Wu J, Liou YC: Regularized methods for the split feasibility problem. Abstr. Appl. Anal. 2012., 2012: Article ID 140679 10.1155/2012/140679Google Scholar
  23. Yu X, Shahzad N, Yao Y: Implicit and explicit algorithms for solving the split feasibility problem. Optim. Lett. 2012, 6: 1447–1462. 10.1007/s11590-011-0340-0View ArticleMathSciNetGoogle Scholar
  24. Zhao J, Yang Q: Self-adaptive projection methods for the multiple-sets split feasibility problem. Inverse Probl. 2011., 27: Article ID 035009 10.1088/0266-5611/27/3/035009Google Scholar
  25. Zhang W, Han D, Li Z: A self-adaptive projection method for solving the multiple-sets split feasibility problem. Inverse Probl. 2009., 25: Article ID 115001 10.1088/0266-5611/25/11/115001Google Scholar
  26. Deimling K: Zeros of accretive operators. Manuscr. Math. 1974, 13: 365–374. 10.1007/BF01171148View ArticleMathSciNetGoogle Scholar
  27. Zhou H: Strong convergence of an explicit iterative algorithm for continuous pseudo-contractions in Banach spaces. Nonlinear Anal. 2009, 70: 4039–4046. 10.1016/j.na.2008.08.012View ArticleMathSciNetGoogle Scholar
  28. Bauschke HH, Combettes PL: A weak-to-strong convergence principle for Fejér-monotone methods in Hilbert spaces. Math. Oper. Res. 2001, 26: 248–264. 10.1287/moor.26.2.248.10558View ArticleMathSciNetGoogle Scholar

Copyright

© Zhu et al.; licensee Springer. 2014

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.