 Research
 Open Access
 Published:
Strong convergence of a relaxed CQ algorithm for the split feasibility problem
Journal of Inequalities and Applications volume 2013, Article number: 197 (2013)
Abstract
The split feasibility problem (SFP) is finding a point in a given closed convex subset of a Hilbert space such that its image under a bounded linear operator belongs to a given closed convex subset of another Hilbert space. The most popular iterative method is Byrne’s CQ algorithm. López et al. proposed a relaxed CQ algorithm for solving SFP where the two closed convex sets are both level sets of convex functions. This algorithm can be implemented easily since it computes projections onto halfspaces and has no need to know a priori the norm of the bounded linear operator. However, their algorithm has only weak convergence in the setting of infinitedimensional Hilbert spaces. In this paper, we introduce a new relaxed CQ algorithm such that the strong convergence is guaranteed. Our result extends and improves the corresponding results of López et al. and some others.
MSC:90C25, 90C30, 47J25.
1 Introduction
The split feasibility problem (SFP) was proposed by Censer and Elfving in [1]. It can mathematically be formulated as the problem of finding a point x satisfying the property:
where A is a given $M\times N$ real matrix (where ${A}^{\ast}$ is the transpose of A), C and Q are nonempty, closed and convex subsets in ${\mathbb{R}}^{N}$ and ${\mathbb{R}}^{M}$, respectively. This problem has received much attention [2] due to its applications in signal processing and image reconstruction, with particular progress in intensitymodulated radiation therapy [3–5], and many other applied fields.
We assume the SFP (1.1) is consistent, and let S be the solution set, i.e.,
It is easily seen that S is closed convex. Many iterative methods can be used to solve the SFP (1.1); see [6–16]. Byrne [6, 17] was among others the first to propose the socalled CQ algorithm, which is defined as follows:
where ${\tau}_{n}\in (0,\frac{2}{{\parallel A\parallel}^{2}})$, and ${P}_{C}$ and ${P}_{Q}$ are the orthogonal projections onto the sets C and Q, respectively. Compared with Censer and Elfving’ algorithm [1], the Byrne’ algorithm is easily executed since it only deal with orthogonal projections with no need to compute matrix inverses.
The CQ algorithm (1.2) can be obtained from optimization. In fact, if we introduce the (convex) objective function
and analyze the minimization problem
then the CQ algorithm (1.2) comes immediately as a special case of the gradient projection algorithm (GPA)(For more details about the GPA, the reader is referred to [18]). Since the convex objective $f(x)$ is differentiable, and has a Lipschitz gradient, which is given by
the GPA for solving the minimization problem (1.4) generates a sequence $({x}_{n})$ recursively as
where the stepsize ${\tau}_{n}$ is chosen in the interval $(0,\frac{2}{L})$, where L is the Lipschitz constant of ∇f.
Observe that in algorithms (1.2) and (1.6) mentioned above, in order to implement the CQ algorithm, one has to compute the operator (matrix) norm $\parallel A\parallel $, which is in general not an easy work in practice. To overcome this difficulty, some authors proposed different adaptive choices of selecting the stepsize ${\tau}_{n}$ (see [6, 14, 19]). For instance, very recently López et al. introduced a new way of selecting the stepsize [19] as follows:
The computation of a projection onto a general closed convex subset is generally difficult. To overcome this difficulty, Fukushima [20] suggested a socalled relaxed projection method to calculate the projection onto a level set of a convex function by computing a sequence of projections onto halfspaces containing the original level set. In the setting of finitedimensional Hilbert spaces, this idea was followed by Yang [13], who introduced the relaxed CQ algorithms for solving SFP (1.1) where the closed convex subsets C and Q are level sets of convex functions.
Recently, for the purpose of generality, the SFP (1.1) is studied in a more general setting. For instance, Xu [12] and López et al. [19] considered the SFP (1.1) in infinitedimensional Hilbert spaces (i.e., the finitedimensional Euclidean spaces ${\mathbb{R}}^{N}$ and ${\mathbb{R}}^{M}$ are replaced with general Hilbert spaces). Very recently, López et al. proposed a relaxed CQ algorithm with a new adaptive way of determining the stepsize sequence $({\tau}_{n})$ for solving the SFP (1.1) where the closed convex subsets C and Q are level sets of convex functions. This algorithm can be implemented easily since it computes projections onto halfspaces and has no need to know a priori the norm of the bounded linear operator. However, their algorithm has only weak convergence in the setting of infinitedimensional Hilbert spaces. In this paper, we introduce a new relaxed CQ algorithm such that the strong convergence is guaranteed in infinitedimensional Hilbert spaces. Our result extends and improves the corresponding results of López et al. and some others.
The rest of this paper is organized as follows. Some useful lemmas are listed in Section 2. In Section 3, the strong convergence of the new relaxed CQ algorithm of this paper is proved.
2 Preliminaries
Throughout the rest of this paper, we denote by H or K a Hilbert space, A is a bounded linear operator from H to K, and by I the identity operator on H or K. If $f:H\to \mathbb{R}$ is a differentiable function, then we denote by ∇f the gradient of the function f. We will also use the notations:

→ denotes strong convergence.

⇀ denotes weak convergence.

${\omega}_{w}({x}_{n})=\{x\mid \mathrm{\exists}\{{x}_{{n}_{k}}\}\subset \{{x}_{n}\}\text{such that}{x}_{{n}_{k}}\rightharpoonup x\}$ denotes the weak ωlimit set of $\{{x}_{n}\}$.
Recall that a mapping $T:H\to H$ is said to be nonexpansive if
$T:H\to H$ is said to be firmly nonexpansive if, for $x,y\in H$,
Recall that a function $f:H\to \mathbb{R}$ is called convex if
A differentiable function f is convex if and only if there holds the inequality:
Recall that an element $g\in H$ is said to be a subgradient of $f:H\to \mathbb{R}$ at x if
This relation is called the subdifferentiable inequality.
A function $f:H\to \mathbb{R}$ is said to be subdifferentiable at x, if it has at least one subgradient at x. The set of subgradients of f at the point x is called the subdifferentiable of f at x, and it is denoted by $\partial f(x)$. A function f is called subdifferentiable, if it is subdifferentiable at all $x\in H$. If a function f is differentiable and convex, then its gradient and subgradient coincide.
A function $f:H\to \mathbb{R}$ is said to be weakly lower semicontinuous (wlsc) at x if ${x}_{n}\rightharpoonup x$ implies
We know that the orthogonal projection ${P}_{C}$ from H onto a nonempty closed convex subset $C\subset H$ is a typical example of a firmly nonexpansive mapping, which is defined by
It is well known that ${P}_{C}$ is characterized by the inequality (for $x\in H$)
The following lemma is not hard to prove (see [17, 21]).
Lemma 2.1 Let f be given as in (1.3). Then

(i)
f is convex and differential.

(ii)
$\mathrm{\nabla}f(x)={A}^{\ast}(I{P}_{Q})Ax$, $x\in H$.

(iii)
f is wlsc on H.

(iv)
∇f is ${\parallel A\parallel}^{2}$Lipschitz: $\parallel \mathrm{\nabla}f(x)\mathrm{\nabla}f(y)\parallel \le {\parallel A\parallel}^{2}\parallel xy\parallel $, $x,y\in H$.
The following are characterizations of firmly nanexpansive mappings (see [22]).
Lemma 2.2 Let $T:H\to H$ be an operator. The following statements are equivalent.

(i)
T is firmly nonexpansive.

(ii)
${\parallel TxTy\parallel}^{2}\le \u3008xy,TxTy\u3009$, $x,y\in H$.

(iii)
$IT$ is firmly nonexpansive.
Lemma 2.3 [23]
Assume $({\alpha}_{n})$ is a sequence of nonnegative real numbers such that
where $({\gamma}_{n})$ is a sequence in $(0,1)$, and $({\sigma}_{n})$ is a sequence in ℝ such that

(i)
${\sum}_{n=0}^{\mathrm{\infty}}{\gamma}_{n}=\mathrm{\infty}$.

(ii)
${lim\hspace{0.17em}sup}_{n\to \mathrm{\infty}}{\sigma}_{n}\le 0$, or ${\sum}_{n=0}^{\mathrm{\infty}}{\gamma}_{n}{\sigma}_{n}<\mathrm{\infty}$.
Then ${lim}_{n\to \mathrm{\infty}}{\alpha}_{n}=0$.
3 Iterative Algorithm
In this section, we turn to consider a new relaxed CQ algorithm in the setting of infinitedimensional Hilbert spaces for solving SFP (1.1) where the closed convex subsets C and Q are of the particular structure, i.e. level sets of convex functions given as follows:
where $c:H\to \mathbb{R}$ and $q:K\to \mathbb{R}$ are convex functions. We assume that both c and q are subdifferentiable on H and K, respectively, and that ∂c and ∂q are bounded operators (i.e., bounded on bounded sets). By the way, we mention that every convex function defined on a finitedimensional Hilbert space is subdifferentiable and its subdifferential operator is a bounded operator (see [24]).
Set
where ${\xi}_{n}\in \partial c({x}_{n})$, and
where ${\zeta}_{n}\in \partial q(A{x}_{n})$.
Obviously, ${C}_{n}$ and ${Q}_{n}$ are halfspaces and it is easy to verify that ${C}_{n}\supset C$ and ${Q}_{n}\supset Q$ hold for every $n\ge 0$ from the subdifferentiable inequality. In what follows, we define
where ${Q}_{n}$ is given as in (3.3). We then have
Firstly, we recall the relaxed CQ algorithm of López et al. [19] for solving the SFP (1.1) where C and Q are given in (3.1) as follows.
Algorithm 3.1 Choose an arbitrary initial guess ${x}_{0}$. Assume ${x}_{n}$ has been constructed. If $\mathrm{\nabla}{f}_{n}({x}_{n})=0$, then stop; otherwise, continue and construct ${x}_{n+1}$ via the formula
where ${C}_{n}$ is given as (3.2), and
López et al. proved that under some certain conditions the sequence $({x}_{n})$ generated by Algorithm 3.1 converges weakly to a solution of the SFP (1.1). Since the projections onto halfspaces ${C}_{n}$ and ${Q}_{n}$ have closed forms and ${\tau}_{n}$ is obtained adaptively via the formula (3.4) (no need to know a priori the norm of operator A), the above relaxed CQ algorithm 3.1 is implementable. But the weak convergence is its a weakness. To overcome this weakness, inspired by Algorithm 3.1, we will introduce a new relaxed CQ algorithm for solving the SFP (1.1) where C and Q are given in (3.1) so that the strong convergence is guaranteed.
It is well known that Halpern’s algorithm has a strong convergence for finding a fixed point of a nonexpansive mapping [25, 26]. Then we are in a position to give our algorithm. The algorithm given below is referred to as a Halperntype algorithm [27].
Algorithm 3.2 Let $u\in H$, and start an initial guess ${x}_{0}\in H$ arbitrarily. Assume that the n th iterate ${x}_{n}$ has been constructed. If $\mathrm{\nabla}{f}_{n}({x}_{n})=0$, then stop (${x}_{n}$ is a approximate solution of SFP (1.1)). Otherwise, continue and calculate the $(n+1)$th iterate ${x}_{n+1}$ via the formula:
where the sequence $({\alpha}_{n})\subset (0,1)$ and $({\tau}_{n})$ and $({\rho}_{n})$ are given as in (3.4).
The convergence result of Algorithm 3.2 is stated in the next theorem.
Theorem 3.3 Assume that $({\alpha}_{n})$ and $({\rho}_{n})$ satisfy the assumptions:
(a1) ${lim}_{n\to \mathrm{\infty}}{\alpha}_{n}=0$ and ${\sum}_{n=0}^{\mathrm{\infty}}{\alpha}_{n}=\mathrm{\infty}$.
(a2) ${inf}_{n}{\rho}_{n}(4{\rho}_{n})>0$.
Then the sequence $({x}_{n})$ generated by Algorithm 3.2 converges in norm to ${P}_{S}u$.
Proof We may assume that the sequence $({x}_{n})$ is infinite, that is, Algorithm 3.2 does not terminate in a finite number of iterations. Thus $\mathrm{\nabla}{f}_{n}({x}_{n})\ne 0$ for all $n\ge 0$. Recall that S is the solution set of the SFP (1.1),
In the consistent case of the SFP (1.1), S is nonempty, closed and convex. Thus, the metric projection ${P}_{S}$ is welldefined. We set $z={P}_{S}u$. Since $z\in S\subset {C}_{n}$ and the projection operator ${P}_{{C}_{n}}$ is nonexpansive, we obtain
Note that $I{P}_{{Q}_{n}}$ is firmly nonexpansive and $\mathrm{\nabla}{f}_{n}(z)=0$, it is deduced from Lemma 2.2 that
which implies that
Thus, we have
Now we prove $({x}_{n})$ is bounded. Indeed, we have from (3.6) that
and consequently
It turns out that
and inductively
and this means that $({x}_{n})$ is bounded. Since ${\alpha}_{n}\to 0$, with no loss of generality, we may assume that there is $\sigma >0$ so that ${\rho}_{n}(4{\rho}_{n})(1{\alpha}_{n})\ge \sigma $ for all n. Setting ${s}_{n}={\parallel {x}_{n}z\parallel}^{2}$, from the last inequality of (3.6), we get the following inequality:
Now, following an idea in [28], we prove ${s}_{n}\to 0$ by distinguishing two cases.
Case 1: $({s}_{n})$ is eventually decreasing (i.e. there exists $k\ge 0$ such that ${s}_{n}>{s}_{n+1}$ holds for all $n\ge k$). In this case, $({s}_{n})$ must be convergent, and from (3.7) it follows that
where $M>0$ is a constant such that $2\parallel {x}_{n+1}z\parallel \parallel uz\parallel \le M$ for all $n\in \mathbb{N}$. Using the condition (a1), we have from (3.8) that $\frac{{f}_{n}^{2}({x}_{n})}{{\parallel \mathrm{\nabla}{f}_{n}({x}_{n})\parallel}^{2}}\to 0$. Thus, to verify that ${f}_{n}({x}_{n})\to 0$, it suffices to show that $(\mathrm{\nabla}f({x}_{n}))$ is bounded. In fact, it follows from Lemma 2.1 that (noting that $\mathrm{\nabla}{f}_{n}(z)=0$ due to $z\in S$)
This implies that $\parallel \mathrm{\nabla}{f}_{n}({x}_{n})\parallel $ is bounded and it yields ${f}_{n}({x}_{n})\to 0$, namely $\parallel (I{P}_{{Q}_{n}})A{x}_{n}\parallel \to 0$.
Since ∂q is bounded on bounded sets, there exists a constant $\eta >0$ such that $\parallel {\zeta}_{n}\parallel \le \eta $ for all $n\ge 0$. From (3.3) and the trivial fact that ${P}_{{Q}_{n}}(A{x}_{n})\in {Q}_{n}$, it follows that
If ${x}^{\ast}\in {\omega}_{w}({x}_{n})$, and $({x}_{{n}_{k}})$ is a subsequence of $({x}_{n})$ such that ${x}_{{n}_{k}}\rightharpoonup {x}^{\ast}$, then the wlsc of q and (3.9) imply that
It turns out that $A{x}^{\ast}\in Q$. Next, we turn to prove ${x}^{\ast}\in C$. For convenience, we set ${y}_{n}:={\alpha}_{n}u+(1{\alpha}_{n})({x}_{n}{\tau}_{n}\mathrm{\nabla}{f}_{n}({x}_{n}))$. In fact, since the ${P}_{{C}_{n}}$ is firmly nonexpansive, it concludes that
On the other hand, we have
and
Noting that $({x}_{n})$ is bounded, we have from (3.10)(3.12) that
where M is some positive constant. Clearly, from (3.13), it turns out that
Thus, we assert that $\parallel (I{P}_{{C}_{n}}){x}_{n}\parallel \to 0$ due to the fact that ${s}_{n}{s}_{n+1}+({\alpha}_{n}+\frac{{f}_{n}({x}_{n})}{\parallel \mathrm{\nabla}{f}_{n}({x}_{n})\parallel})M\to 0$. Moreover, by the definition of ${C}_{n}$, we obtain
where δ is a constant such that $\parallel {\xi}_{n}\parallel \le \delta $ for all $n\ge 0$. The wlsc of c then implies that
Consequently, ${x}^{\ast}\in C$, and hence ${\omega}_{w}({x}_{n})\subset S$. Furthermore, due to (2.2), we get
Taking into account of (3.7), we have
Applying Lemma 2.3 to (3.16), we obtain ${s}_{n}\to 0$.
Case 2: $({s}_{n})$ is not eventually decreasing, that is, we can find an integer ${n}_{0}$ such that ${s}_{{n}_{0}}\le {s}_{{n}_{0}+1}$. Now we define
It is easy to see that ${V}_{n}$ is nonempty and satisfies ${V}_{n}\subseteq {V}_{n+1}$. Let
It is clear that $\psi (n)\to \mathrm{\infty}$ as $n\to \mathrm{\infty}$ (otherwise, $({s}_{n})$ is eventually decreasing). It is also clear that ${s}_{\psi (n)}\le {s}_{\psi (n)+1}$ for all $n>{n}_{0}$. Moreover,
In fact, if $\psi (n)=n$, then the inequity (3.17) is trivial; if $\psi (n)<n$, from the definition of $\psi (n)$, there exists some $i\in \mathbb{N}$ such that $\psi (n)+i=n$, we deduce that
and the inequity (3.17) holds again. Since ${s}_{\psi (n)}\le {s}_{\psi (n)+1}$ for all $n>{n}_{0}$, it follows from (3.8) that
so that ${f}_{\psi (n)}({x}_{\psi (n)})\to 0$ as $n\to \mathrm{\infty}$ (noting that $\parallel \mathrm{\nabla}{f}_{\psi (n)}({x}_{\psi (n)})\parallel $ is bounded). By the same argument to the proof in case 1, we have ${\omega}_{w}({x}_{\psi (n)})\subset S$. On the other hand, noting ${s}_{\psi (n)}\le {s}_{\psi (n)+1}$ again, we have from (3.12) and (3.14) that
where M is a positive constant. Letting $n\to \mathrm{\infty}$ yields that
from which one can deduce that
Since ${s}_{\psi (n)}\le {s}_{\psi (n)+1}$, it follows from (3.7) that
Combining (3.19) and (3.20) yields
and hence ${s}_{\psi (n)}\to 0$, which together with (3.18) implies that
which, together with (3.17), in turn implies that ${s}_{n}\to 0$, that is, ${x}_{n}\to z$. □
Remark 3.4 Since u can be chosen in H arbitrarily, one can compute the minimumnorm solution of SFP (1.1) where C and Q are given in (3.1) by taking $u=0$ in Algorithm 3.2 whether $0\in C$ or $0\notin C$.
References
 1.
Censor Y, Elfving T: A multiprojection algorithm using Bregman projection in product space. Numer. Algorithms 1994, 8(2–4):221–239.
 2.
López G, MartínMárquez V, Xu HK: Iterative algorithms for the multipulsets split feasiblity 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, WI; 2010:243–279.
 3.
Censor Y, Bortfeld T, Martin B, Trofimov A: A unified approach for inversion problems in intensitymodulated radiation therapy. Phys. Med. Biol. 2003, 51: 2353–2365.
 4.
Censor Y, Elfving T, Kopf N, Bortfeld T: The multiplesets split feasibility problem and its applications for inverse problem. Inverse Probl. 2005, 21: 2071–2084. 10.1088/02665611/21/6/017
 5.
López G, MartínMárquez V, Xu HK: Perturbation techniques for nonexpansive mappings with applications. Nonlinear Anal., Real World Appl. 2009, 10: 2369–2383. 10.1016/j.nonrwa.2008.04.020
 6.
Byrne C: Iterative oblique projection onto convex sets and the split feasibity problem. Inverse Probl. 2002, 18: 441–453. 10.1088/02665611/18/2/310
 7.
Dang Y, Gao Y: The strong convergence of a KMCQlike algorithm for a split feasibility problem. Inverse Probl. 2011., 27: Article ID 015007
 8.
Qu B, Xiu N: A note on the CQ algorithm for the split feasibility problem. Inverse Probl. 2005, 21: 1655–1665. 10.1088/02665611/21/5/009
 9.
Schöpfer F, Schuster T, Louis AK: An iterative regularization method for the solution of the split feasibility problem in Banach spaces. Inverse Probl. 2008., 24: Article ID 055008
 10.
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
 11.
Xu HK: A variable Krasonosel’skiiMann algorithm and the multipleset split feasibility problem. Inverse Probl. 2006, 22: 2021–2034. 10.1088/02665611/22/6/007
 12.
Xu HK: Iterative methods for the split feasibility problem in infinitedimensional Hilbert spaces. Inverse Probl. 2010., 26: Article ID 105018
 13.
Yang Q: The relaxed CQ algorithm for solving the split feasibility problem. Inverse Probl. 2004, 20: 1261–1266. 10.1088/02665611/20/4/014
 14.
Yang Q: On variableset relaxed projection algorithm for variational inequalities. J. Math. Anal. Appl. 2005, 302: 166–179. 10.1016/j.jmaa.2004.07.048
 15.
Zhao J, Yang Q: Generalized KM theorems and their applications. Inverse Probl. 2006, 22(3):833–844. 10.1088/02665611/22/3/006
 16.
Zhao J, Yang Q: Selfadaptive projection methods for the multiplesets split feasibility problem. Inverse Probl. 2011., 27: Article ID 035009
 17.
Byrne C: A unified treatment of some iterative algorithms in signal processing and image reconstruction. Inverse Probl. 2004, 20(1):103–120. 10.1088/02665611/20/1/006
 18.
Figueiredo MA, Nowak RD, Wright SJ: Gradient projection for space reconstruction: application to compressed sensing and other inverse problems. IEEE J. Sel. Top. Signal Process. 2007, 1: 586–598.
 19.
López G, MartínMárquez V, Wang FH, Xu HK: Solving the split feasibility problem without prior knowledge of matrix norms. Inverse Probl. 2012. doi:10.1088/0266–5611/28/8/085004
 20.
Fukushima M: A relaxed projection method for variational inequalities. Math. Program. 1986, 35: 58–70. 10.1007/BF01589441
 21.
Aubin JP: Optima and Equilibria: An Introduction to Nonlinear Analysis. Springer, Berlin; 1993.
 22.
Goebel K, Kirk WA: Topics on Metric Fixed Point Theory. Cambridge University Press, Cambridge; 1990.
 23.
Xu HK: Iterative algorithms for nonlinnear operators. J. Lond. Math. Soc. 2002, 66: 240–256. 10.1112/S0024610702003332
 24.
Bauschke HH, Borwein JM: On projection algorithms for solving convex feasibility problem. SIAM Rev. 1996, 38: 367–426. 10.1137/S0036144593251710
 25.
Suzuki T: A sufficient and necessary condition for Halperntype strong convergence to fixed points of nonexpansive mappings. Proc. Am. Math. Soc. 2007, 135: 99–106.
 26.
Xu HK: Viscosity approximation methods for nonexpansive mappings. J. Math. Anal. Appl. 2004, 298: 279–291. 10.1016/j.jmaa.2004.04.059
 27.
Halpern B: Fixed points of nonexpanding maps. Bull. Am. Math. Soc. 1967, 73: 957–961. 10.1090/S000299041967118640
 28.
Maingé PE: A hybrid extragradientviscosity method for monotone operators and fixed point problems. SIAM J. Control Optim. 2008, 47: 1499–1515. 10.1137/060675319
Acknowledgements
The authors wish to thank the referees for their helpful comments, which notably improved the presentation of this manuscript. This work was supported by the Fundamental Research Funds for the Central Universities (ZXH2012K001) and in part by the Foundation of Tianjin Key Lab for Advanced Signal Processing.
Author information
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All the authors read and approved the final manuscript.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
About this article
Received
Accepted
Published
DOI
Keywords
 split feasibility problem
 relaxed CQ algorithm
 Hilbert space
 strong convergence
 bounded linear operator