nach oben

Journal of Inequalities and Applications

Erschienen in:

Open Access 01.12.2013 | Research

Split feasibility and fixed-point problems for asymptotically quasi-nonexpansive mappings

verfasst von: Jitsupa Deepho, Poom Kumam

Erschienen in: Journal of Inequalities and Applications | Ausgabe 1/2013

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config

KI-gestützte Suche

Patentsuche

Aus

Abstract

The purpose of this paper is to introduce and analyze a weakly convergent theorem by using the regularized method and the relaxed extragradient method for finding a common element of the solution set Γ of the split feasibility problem and

Fix (T)

of fixed points of asymptotically quasi-nonexpansive mappings T in the setting of infinite-dimensional Hilbert spaces. Consequently, we prove that the sequence generated by the proposed algorithm converges weakly to an element of

Fix (T) \cap Γ

under mild assumptions.

MSC:47H09, 47J25, 65K10.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

All authors contributed equally and significantly in writing this paper. All authors read and approved the final manuscript.

1 Introduction

In 1994, Censor and Elfving [1] first introduced the split feasibility problem (SFP) in finite-dimensional Hilbert spaces for modeling inverse problems which arise from phase retrievals and in medical image reconstruction [2]. It was found that the SFP can also be used to model intensity-modulated radiation therapy (IMRT) (see [3‐6]). Very recently, Xu [7] considered the SFP in the framework of infinite-dimensional Hilbert spaces. In this setting, the SFP is formulated as the problem of finding a point

x^{*}

with the property

x^{*} \in C and A x^{*} \in Q,

(1.1)

where C and Q are the nonempty closed convex subsets of the infinite-dimensional real Hilbert spaces

H_{1}

and

H_{2}

, respectively. Let

A \in B (H_{1}, H_{2})

, where

B (H_{1}, H_{2})

denotes the family of all bounded linear operators from

H_{1}

H_{2}

We use Γ to denote the solution set of the SFP, i.e.,

Γ = {x \in C : A x \in Q} .

Assume that the SFP is consistent (i.e., (1.1) has a solution) so that Γ is closed, convex and nonempty. A special case of the SFP is the following convex constrained linear inverse problem:

find x \in C such that A x = b,

(1.2)

which has extensively been investigated by using the Landweber iterative method [8]:

\begin{matrix} let x_{0} be arbitrary for n = 0, 1, \dots, let \\ x_{n + 1} = x_{n} + γ A^{T} (b - A x_{n}) . \end{matrix}

Comparatively, the SFP has received much less attention so far due to the complexity resulting from the set Q. Therefore, whether various versions of the projected Landweber iterative method [8] can be extended to solve the SFP remains an interesting open topic.

The original algorithm given in [1] involves the computation of the inverse

A^{- 1}

(assuming the existence of the inverse of A):

x_{k + 1} = A^{- 1} P_{Q} (P_{A (C)} (A x_{k})), k \geq 0,

where

C, Q \subset R^{n}

are closed convex sets, A is a full rank

n \times n

matrix and

A (C) = {y \in R^{n} | y = A x, x \in C}

, and thus has not become popular.

A more popular algorithm that solves the SFP seems to be the CQ algorithm of Byrne [2, 9] which is found to be a gradient-projection method (GPM) in convex minimization. It is also a special case of the proximal forward-backward splitting method [10]. The CQ algorithm only involves the computations of the projections

P_{C}

and

P_{Q}

onto the sets C and Q, respectively, and is therefore implementable in the case where

P_{C}

and

P_{Q}

have closed-form expressions (for example, C and Q are closed balls or half-spaces). It remains, however, a challenge on the CQ algorithm in the case where the projection

P_{C}

and/or

P_{Q}

fail to have closed-form expressions though theoretically we can prove the (weak) convergence of the algorithm.

Recently, Xu [7] gave a continuation of the study on the CQ algorithm and its convergence. He applied Mann’s algorithm to the SFP and proposed an averaged CQ algorithm, which was proved to be weakly convergent to a solution of the SFP. He derived a weak convergence result, which shows that for suitable choices of iterative parameters (including the regularization), the sequence of iterative solutions can converge weakly to an exact solution of the SFP. He also established the strong convergence result, which shows that the minimum-norm solution can be obtained. Later, Deepho and Kumam [11] extended the results of Xu [7] by introducing and studying the modified Halpern iterative scheme for solving the split feasibility problem (SFP) in the setting of infinite-dimensional Hilbert spaces.

Throughout this paper, we always assume that the SFP is consistent, that is, the solution set Γ of the SFP is nonempty. Let

f : H_{1} \to R

be a continuous differentiable function. The minimization problem

min_{x \in C} f (x) : = \frac{1}{2} {∥ A x - P_{Q} A x ∥}^{2}

(1.3)

is ill-posed. Therefore (see [7]), consider the following Tikhonov regularized problem:

min_{x \in C} f_{α} (x) : = \frac{1}{2} {∥ A x - P_{Q} A x ∥}^{2} + \frac{1}{2} α {∥ x ∥}^{2},

(1.4)

where

α > 0

is the regularization parameter.

We observe that the gradient

\nabla f_{α} (x) = \nabla f (x) + α I = A^{*} (I - P_{Q}) A + α I

(1.5)

(α + {∥ A ∥}^{2})

-Lipschitz continuous and α-strongly monotone.

Define the Picard iterates

x_{n + 1}^{α} = P_{C} (I - γ (A^{*} (I - P_{Q}) A + α I)) x_{n}^{α} .

(1.6)

Xu [7] showed that if SFP (1.1) is consistent, then as

n \to \infty

x_{n}^{α} \to x_{α}

and consequently the strong

{lim}_{α \to 0} x_{α}

exists and is the minimum-norm solution of the SFP. Note that (1.6) is double-step iteration. Xu [7] further suggested the following single step regularized method:

x_{n + 1} = P_{C} (I - γ \nabla f_{α_{n}}) x_{n} = P_{C} ((1 - α_{n} γ_{n}) x_{n} - γ_{n} A^{*} (I - P_{Q}) A x_{n}) .

(1.7)

He proved that the sequence

{x_{n}}

generated by (1.7) converges in norm to the minimum-norm solution of the SFP provided the parameters

{α_{n}}

and

{γ_{n}}

satisfy the following conditions:

(i)

α_{n} \to 0

and

0 < γ_{n} < \frac{α_{n}}{{∥ A ∥}^{2} + α_{n}}

;

(ii)

\sum_{n} α_{n} γ_{n} = \infty

;

(iii)

\frac{| γ_{n + 1} - γ_{n} | + γ_{n} | α_{n + 1} - α_{n} |}{{(α_{n + 1} γ_{n + 1})}^{2}} \to 0

Motivated by the idea of the relaxed extragradient method and Xu’s regularization, Ceng, Ansari and Yao [12] presented the following relaxed extragradient method with regularization for finding a common element of the solution set of the split feasibility problem and the set

Fix (S)

of fixed points of a nonexpansive mapping S:

{\begin{matrix} x_{0} = x \in C chosen arbitrarily, \\ y_{n} = (1 - β_{n}) + β_{n} P_{C} (x_{n} - λ \nabla f_{α_{n}} (x_{n})), \\ x_{n + 1} = γ_{n} x_{n} + (1 - γ_{n}) S P_{C} (y_{n} - λ \nabla f_{α_{n}} (y_{n})), \forall n \geq 0 . \end{matrix}

(1.8)

They only obtained the weak convergence of iterative algorithm (1.8).

The purpose of this paper to study and analyze an relaxed extragradient method with regularization for finding a common element of the solution set Γ of the SFP and the set solutions of fixed points for asymptotically quasi-nonexpansive mappings and a Lipschitz continuous mapping in a real Hilbert space. We prove that the sequence generated by the proposed method converges weakly to an element

\hat{x}

Fix (T) \cap Γ

2 Preliminaries

We first recall some definitions, notations, and conclusions which will be needed in proving our main results. Let H be a real Hilbert space with the inner product

〈 \cdot, \cdot 〉

and

∥ \cdot ∥

and let C be a closed and convex subset of H. Let E be a Banach space. A mapping

T : E \to E

is said to be demi-closed at origin if for any sequences

{x_{n}} \subset E

with

x_{n} ⇀ x^{*}

and

∥ (I - T) x_{n} ∥ \to 0

x^{*} = T x^{*}

. A Banach space E is said to have the Opial property if for any sequence

{x_{n}}

with

x_{n} ⇀ x^{*}

\underset{n \to \infty}{lim inf} ∥ x_{n} - x^{*} ∥ < \underset{n \to \infty}{lim inf} ∥ x_{n} - y ∥, \forall y \in E with y \neq x^{*} .

Remark 2.1 It is well known that each Hilbert space possesses the Opial property.

Definition 2.2 Let H be a real Hilbert space, let C be a nonempty and closed convex subset. We denote by

Fix (T)

the set of fixed points of T, that is,

Fix (T) = {x \in C : x = T x}

. Then T is said to be

(i)

nonexpansive if

∥ T x - T y ∥ \leq ∥ x - y ∥

for all

x, y \in C

;

(ii)

quasi-nonexpansive if

∥ T x - p ∥ \leq ∥ x - p ∥

for all

x \in C

and

p \in F (T)

;

(iii)

asymptotically nonexpansive if there exist a sequence

k_{n} \geq 1

and

{lim}_{n \to \infty} k_{n} = 1

such that

∥ T^{n} x - T^{n} y ∥ \leq k_{n} ∥ x - y ∥

for all

x, y \in C

and

n \geq 1

;

(iv)

asymptotically quasi-nonexpansive if there exist a sequence

k_{n} \geq 1

and

{lim}_{n \to \infty} k_{n} = 1

such that

∥ T^{n} x - p ∥ \leq k_{n} ∥ x - p ∥

for all

x \in C

p \in F (T)

and

n \geq 1

;

(v)

uniformlyL-Lipschitzian if there exists a constant

L > 0

such that

∥ T^{n} x - T^{n} y ∥ \leq L ∥ x - y ∥

for all

x, y \in C

and

n \geq 1

Remark 2.3 By the above definitions, it is clear that:

(i)

a nonexpansive mapping is an asymptotically quasi-nonexpansive mapping;

(ii)

a quasi-nonexpansive mapping is an asymptotically-quasi nonexpansive mapping;

(iii)

an asymptotically nonexpansive mapping is an asymptotically quasi-nonexpansive mapping.

Proposition 2.4 (see [9])

We have the following assertions.

(i)

Tis nonexpansive if and only if the complement

I - T

\frac{1}{2}

-ism.

(ii)

IfTisν-ism and

γ > 0

, thenγTis

\frac{ν}{γ}

-ism.

(iii)

Tis averaged if and only if the complement

I - T

isν-ism for some

ν > \frac{1}{2}

Indeed, for

α \in (0, 1)

, Tisα-averaged if and only if

I - T

\frac{1}{2 α}

-ism.

Proposition 2.5 (see [9, 13])

We have the following assertions.

(i)

T = (1 - α) S + α V

for some

α \in (0, 1)

, Sis averaged andVis nonexpansive, thenTis averaged.

(ii)

Tis firmly nonexpansive if and only if the complement

I - T

is firmly nonexpansive.

(iii)

T = (1 - α) S + α V

for some

α \in (0, 1)

, Sis firmly nonexpansive andVis nonexpansive, thenTis averaged.

(iv)

The composite of finite many averaged mappings is averaged. That is, if each of the mappings

{T_{i}}_{i = 1}^{n}

is averaged, then so is the composite

T_{1} \circ T_{2} \circ \dots \circ T_{N}

. In particular, if

T_{1}

α_{1}

-averaged and

T_{2}

α_{2}

-averaged, where

α_{1}, α_{2} \in (0, 1)

, then the composite

T_{1} \circ T_{2}

isα-averaged, where

α = α_{1} + α_{2} - α_{1} α_{2}

(v)

If the mappings

{T_{i}}_{i = 1}^{n}

are averaged and have a common fixed point, then

⋂_{i = 1}^{n} Fix (T_{i}) = Fix (T_{1} \dots T_{N}) .

Lemma 2.6 (see [14], demiclosedness principle)

LetCbe a nonempty closed and convex subset of a real Hilbert spaceHand let

S : C \to C

be a nonexpansive mapping with

Fix (S) \neq \emptyset

. If the sequence

{x_{n}} \subseteq C

converges weakly toxand the sequence

{(I - S) x_{n}}

converges strongly toy, then

(I - S) x = y

; in particular, if

y = 0

, then

x \in Fix (S)

Lemma 2.7 (see [15])

Let the sequences

{a_{n}}

and

{u_{n}}

of real numbers satisfy

a_{n + 1} \leq (1 + u_{n}) a_{n}, \forall n \geq 1,

where

a_{n} \geq 0

u_{n} \geq 0

and

\sum_{n = 1}^{\infty} u_{n} < \infty

. Then

(1)

{lim}_{n \to \infty} a_{n}

exists;

(2)

{lim inf}_{n \to \infty} a_{n} = 0

, then

{lim}_{n \to \infty} a_{n} = 0

The following lemma gives some characterizations and useful properties of the metric projection

P_{C}

in a Hilbert space.

For every point

x \in H

, there exists a unique nearest point in C, denoted by

P_{C} x

, such that

∥ x - P_{C} x ∥ \leq ∥ x - y ∥, \forall y \in C,

(2.1)

where

P_{C}

is called the metric projection ofH onto C. We know that

P_{C}

is a nonexpansive mapping of H onto C.

Proposition 2.8For given

x \in H

and

z \in C

(i)

z = P_{C} x

if and only if

〈 x - z, y - z 〉 \leq 0

for all

y \in C

(ii)

z = P_{C} x

if and only if

{∥ x - z ∥}^{2} \leq {∥ x - y ∥}^{2} - {∥ y - z ∥}^{2}

for all

y \in C

(iii)

For all

y \in H

〈 P_{C} x - P_{C} y, x - y 〉 \geq {∥ P_{C} x - P_{C} y ∥}^{2}

Lemma 2.9 (see [16])

LetHbe a real Hilbert space. Then the following equations hold:

(i)

{∥ x - y ∥}^{2} = {∥ x ∥}^{2} - {∥ y ∥}^{2} - 2 〈 x - y, y 〉

for all

x, y \in H

;

(ii)

{∥ t x + (1 - t) y ∥}^{2} = t {∥ x ∥}^{2} + (1 - t) {∥ y ∥}^{2} - t (1 - t) {∥ x - y ∥}^{2}

for all

t \in [0, 1]

and

x, y \in H

Let K be a nonempty closed convex subset of a real Hilbert space H and let

F : K \to H

be a monotone mapping. The variational inequality problem (VIP) is to find

x \in K

such that

〈 F x, y - x 〉 \geq 0, \forall y \in K .

The solution set of the VIP is denoted by

VIP (K, F)

. It is well known that

x \in VI (K, F) \Leftrightarrow x = P_{K} (x - λ F x), \forall λ > 0 .

A set-valued mapping

T : H \to 2^{H}

is called monotone if for all

x, y \in H

f \in T x

and

g \in T y

imply

〈 x - y, f - g 〉 \geq 0 .

A monotone mapping

T : H \to 2^{H}

is called maximal if its graph

G (T)

is not properly contained in the graph of any other monotone mapping. It is known that a monotone mapping T is maximal if and only if for

(x, f) \in H \times H

〈 x - y, f - g 〉 \geq 0

for every

(y, g) \in G (T)

implies

f \in T x

. Let

F : K \to H

be a monotone and k-Lipschitz continuous mapping and let

N_{K} v

be the normal cone to K at

v \in K

, that is,

N_{K} v = {w \in H : 〈 v - u, w 〉 \geq 0, \forall u \in K} .

Define

T v = {\begin{matrix} F v + N_{K} v & if v \in K, \\ \emptyset & if v \notin K . \end{matrix}

Then T is maximal monotone and

0 \in T v

if and only if

v \in VI (K, F)

; see [15] for more details.

We can use fixed point algorithms to solve the SFP on the basis of the following observation.

Let

λ > 0

and assume that

x^{*} \in Γ

. Then

A x^{*} \in Q

, which implies that

(I - P_{Q}) A x^{*} = 0

, and thus

λ A^{*} (I - P_{Q}) A x^{*} = 0

. Hence, we have the fixed point equation

(I - λ A^{*} (I - P_{Q}) A) x^{*} = x^{*}

. Requiring that

x^{*} \in C

, we consider the fixed point equation

P_{C} (I - λ \nabla f) x^{*} = P_{C} (I - λ A^{*} (I - P_{Q}) A) x^{*} = x^{*} .

(2.2)

It is proved in [[7], Proposition 3.2] that the solutions of fixed point equation (2.2) are exactly the solutions of the SFP; namely, for given

x^{*} \in H_{1}

x^{*}

solves the SFP if and only if

x^{*}

solves fixed point equation (2.2).

Proposition 2.10 (see [12])

Given

x^{*} \in H_{1}

, the following statements are equivalent.

(i)

x^{*}

solves the SFP;

(ii)

x^{*}

solves fixed point equation (2.2);

(iii)

x^{*}

solves the variational inequality problem (VIP) of finding

x^{*} \in C

such that

〈 \nabla f (x^{*}), x - x^{*} 〉 \geq 0, \forall x \in C,

(2.3)

where

\nabla f = A^{*} (I - P_{Q}) A

and

A^{*}

is the adjoint ofA.

Proof (i) ⇔ (ii). See the proof in [[7], Proposition 3.2].

(ii)

⇔ (iii). Observe that

\begin{array}{rcl} P_{C} (I - λ A^{*} (I - P_{Q}) A) x^{*} = x^{*} & \Leftrightarrow & 〈 (I - λ A^{*} (I - P_{Q}) A) x^{*} - x^{*}, x - x^{*} 〉 \leq 0, \forall x \in C \\ \Leftrightarrow & - λ 〈 A^{*} (I - P_{Q}) A x^{*}, x - x^{*} 〉 \leq 0, \forall x \in C \\ \Leftrightarrow & 〈 \nabla f (x^{*}), x - x^{*} 〉 \geq 0, \forall x \in C, \end{array}

where

\nabla f = A^{*} (I - P_{Q}) A

. □

Remark 2.11 It is clear from Proposition 2.10 that

Γ : = Fix (P_{C} (I - λ \nabla f)) = VI (C, \nabla f)

for any

λ > 0

, where

Fix (P_{C} (I - λ \nabla f))

and

VI (C, \nabla f)

denote the set of fixed points of

P_{C} (I - λ \nabla f)

and the solution set of VIP.

3 Main result

Theorem 3.1LetCbe a nonempty, closed, and convex subset of a real Hilbert spaceHand let

T : C \to C

be a uniformlyL-Lipschitzian and asymptotically quasi-nonexpansive mappings with

Fix (T) \cap Γ \neq \emptyset

and

{k_{n}} \subset [1, \infty)

for all

n \in N

such that

\sum_{n = 1}^{\infty} (k_{n} - 1) < \infty

. Let

{x_{n}}

and

{y_{n}}

be the sequences inCgenerated by the following algorithm:

{\begin{matrix} x_{0} = x \in C chosen arbitrarily, \\ y_{n} = P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n}, \\ x_{n + 1} = β_{n} x_{n} + (1 - β_{n}) T^{n} y_{n}, \forall n \geq 0, \end{matrix}

(3.1)

where

\nabla f_{α_{n}} = \nabla f + α_{n} I = A^{*} (I - P_{Q}) A + α_{n} I

, and three sequences

{α_{n}}

{λ_{n}}

, and

{β_{n}}

satisfy the conditions:

(i)

\sum_{n = 1}^{\infty} α_{n} < \infty

(ii)

{λ_{n}} \subset [a, b]

for some

a, b \in (0, \frac{1}{{∥ A ∥}^{2}})

and

\sum_{n = 1}^{\infty} | λ_{n + 1} - λ_{n} | < \infty

(iii)

{β_{n}} \subset [c, d]

for some

c, d \in (0, 1)

Then the sequences

{x_{n}}

and

{y_{n}}

converge weakly to an element

\hat{x} \in Fix (T) \cap Γ

Proof We first show that

P_{C} (I - λ \nabla f_{α})

is ζ-averaged for each

λ_{n} \in (0, \frac{2}{α + {∥ A ∥}^{2}})

, where

ζ = \frac{2 + λ (α + {∥ A ∥}^{2})}{4} .

Indeed, it is easy to see that

\nabla f = A^{*} (I - P_{Q}) A

\frac{1}{{∥ A ∥}^{2}}

-ism, that is,

〈 \nabla f (x) - \nabla f (y), x - y 〉 \geq \frac{1}{{∥ A ∥}^{2}} {∥ \nabla f (x) - \nabla f (y) ∥}^{2} .

Observe that

\begin{matrix} (α + {∥ A ∥}^{2}) 〈 \nabla f_{α} (x) - \nabla f_{α} (y), x - y 〉 \\ = (α + {∥ A ∥}^{2}) [α {∥ x - y ∥}^{2} + 〈 \nabla f (x) - \nabla f (y), x - y 〉] \\ = α^{2} {∥ x - y ∥}^{2} + α 〈 \nabla f (x) - \nabla f (y), x - y 〉 \\ + α {∥ A ∥}^{2} {∥ x - y ∥}^{2} + {∥ A ∥}^{2} 〈 \nabla f (x) - \nabla f (y), x - y 〉 \\ \geq α^{2} {∥ x - y ∥}^{2} + 2 α 〈 \nabla f (x) - \nabla f (y), x - y 〉 + {∥ \nabla f (x) - \nabla f (y) ∥}^{2} \\ = {∥ α (x - y) + \nabla f (x) - \nabla f (y) ∥}^{2} \\ = {∥ \nabla f (x) - \nabla f (y) ∥}^{2} . \end{matrix}

Hence, it follows that

\nabla f_{α} = α I + A^{*} (I - P_{Q}) A

\frac{1}{α + {∥ A ∥}^{2}}

-ism. Thus,

λ \nabla f_{α}

\frac{1}{λ (α + {∥ A ∥}^{2})}

-ism. By Proposition 2.4(iii) the composite

(I - λ \nabla f_{α})

\frac{λ (α + {∥ A ∥}^{2})}{2}

-averaged. Therefore, noting that

P_{C}

\frac{1}{2}

-averaged and utilizing Proposition 2.5(iv), we know that for each

λ \in (0, \frac{2}{α + {∥ A ∥}^{2}})

P_{C} (I - λ \nabla f_{α})

is ζ-averaged with

ζ = \frac{1}{2} + \frac{λ (α + {∥ A ∥}^{2})}{2} - \frac{1}{2} \cdot \frac{λ (α + {∥ A ∥}^{2})}{2} = \frac{2 + λ (α + {∥ A ∥}^{2})}{4} \in (0, 1) .

This shows that

P_{C} (I - λ \nabla f_{α})

is nonexpansive. Furthermore, for

{λ_{n}} \in [a, b]

with

a, b \in (0, \frac{1}{{∥ A ∥}^{2}})

, utilizing the fact that

{lim}_{n \to \infty} \frac{1}{α_{n} + {∥ A ∥}^{2}} = \frac{1}{{∥ A ∥}^{2}}

, we may assume that

0 < a \leq λ_{n} \leq b < \frac{1}{{∥ A ∥}^{2}}, \forall n \geq 0 .

Consequently, it follows that for each integer

n \geq 0

P_{C} (I - λ_{n} \nabla f_{α_{n}})

ζ_{n}

-averaged with

ζ_{n} = \frac{1}{2} + \frac{λ_{n} (α_{n} + {∥ A ∥}^{2})}{2} - \frac{1}{2} \cdot \frac{λ_{n} (α_{n} + {∥ A ∥}^{2})}{2} = \frac{2 + λ_{n} (α_{n} + {∥ A ∥}^{2})}{4} \in (0, 1) .

This immediately implies that

P_{C} (I - λ_{n} \nabla f_{α_{n}})

is nonexpansive for all

n \geq 0

We divide the remainder of the proof into several steps.

Step 1. We prove that

{x_{n}}

is bounded. Indeed, we take a fixed

p \in Fix (T) \cap Γ

arbitrarily. Then we get

P_{C} (I - λ_{n} \nabla f) p = p

for

λ_{n} \in (0, \frac{2}{{∥ A ∥}^{2}})

. Since

P_{C}

and

(I - λ_{n} \nabla f_{α_{n}})

are nonexpansive mappings, then we have

\begin{array}{rcl} ∥ y_{n} - p ∥ & = & ∥ P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n} - P_{C} (I - λ_{n} \nabla f) p ∥ \\ \leq & ∥ P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n} - P_{C} (I - λ_{n} \nabla f_{α_{n}}) p ∥ \\ + ∥ P_{C} (I - λ_{n} \nabla f_{α_{n}}) p - P_{C} (I - λ_{n} \nabla f) p ∥ \\ \leq & ∥ x_{n} - p ∥ + ∥ (I - λ_{n} \nabla f_{α_{n}}) p - (I - λ_{n} \nabla f) p ∥ \\ = & ∥ x_{n} - p ∥ + ∥ λ_{n} \nabla f p - λ_{n} \nabla f_{α_{n}} p ∥ \\ = & ∥ x_{n} - p ∥ + λ_{n} ∥ \nabla f p - \nabla f_{α_{n}} p ∥ \\ = & ∥ x_{n} - p ∥ + λ_{n} ∥ \nabla f p - \nabla f p - α_{n} p ∥ \\ \leq & ∥ x_{n} - p ∥ + α_{n} λ_{n} ∥ p ∥ . \end{array}

(3.2)

Observe that

\begin{array}{rcl} ∥ x_{n + 1} - p ∥ & = & ∥ β_{n} x_{n} + (1 - β_{n}) T^{n} y_{n} - p ∥ \\ \leq & β_{n} ∥ x_{n} - p ∥ + (1 - β_{n}) ∥ T^{n} y_{n} - p ∥ \\ \leq & β_{n} ∥ x_{n} - p ∥ + (1 - β_{n}) k_{n} ∥ y_{n} - p ∥ \\ \leq & β_{n} ∥ x_{n} - p ∥ + (1 - β_{n}) k_{n} (∥ x_{n} - p ∥ + λ_{n} α_{n} ∥ p ∥) \\ = & β_{n} ∥ x_{n} - p ∥ + (1 - β_{n}) k_{n} ∥ x_{n} - p ∥ + (1 - β_{n}) k_{n} α_{n} λ_{n} ∥ p ∥ \\ = & (1 + (k_{n} - 1) (1 - β_{n})) ∥ x_{n} - p ∥ + (1 - β_{n}) k_{n} α_{n} λ_{n} ∥ p ∥ . \end{array}

(3.3)

Since

\sum_{n = 1}^{\infty} (k_{n} - 1) < \infty

, according to Lemma 2.7 and (i), (ii) and (3.3), we obtain that

lim_{n \to \infty} ∥ x_{n} - p ∥ exists for each p \in Fix (T) \cap Γ .

(3.4)

This implies that

{x_{n}}

is bounded and

{y_{n}}

is also bounded.

It follows that

∥ T^{n} x_{n} - p ∥ \leq k_{n} ∥ x_{n} - p ∥ .

Hence

{T^{n} x_{n} - p}

is bounded.

Step 2. We prove that

lim_{n \to \infty} ∥ y_{n} - T y_{n} ∥ = 0 .

In fact, it follows from (3.2) that

\begin{array}{rcl} {∥ y_{n} - p ∥}^{2} & = & {(∥ x_{n} - p ∥ + α_{n} λ_{n} ∥ p ∥)}^{2} \\ \leq & {∥ x_{n} - p ∥}^{2} + 2 α_{n} λ_{n} ∥ p ∥ ∥ x_{n} - p ∥ + α_{n}^{2} λ_{n}^{2} {∥ p ∥}^{2} \\ = & {∥ x_{n} - p ∥}^{2} + α_{n} (2 λ_{n} ∥ p ∥ ∥ x_{n} - p ∥ + α_{n} λ_{n}^{2} {∥ p ∥}^{2}) \\ = & {∥ x_{n} - p ∥}^{2} + α_{n} M, \end{array}

where

M = {sup}_{n \geq 0} {2 λ_{n} ∥ p ∥ ∥ x_{n} - p ∥ + α_{n} λ_{n}^{2} {∥ p ∥}^{2}} < \infty

It follows that

\begin{array}{rcl} {∥ T^{n} y_{n} - p ∥}^{2} & \leq & {(k_{n} ∥ y_{n} - p ∥)}^{2} \\ = & k_{n}^{2} {∥ y_{n} - p ∥}^{2} \\ = & k_{n}^{2} {∥ x_{n} - p ∥}^{2} + α_{n} k_{n}^{2} M . \end{array}

Also, observe that

\begin{array}{rcl} {∥ x_{n + 1} - p ∥}^{2} & = & {∥ β_{n} x_{n} + (1 - β_{n}) T^{n} y_{n} - p ∥}^{2} \\ \leq & β_{n} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) {∥ T^{n} y_{n} - p ∥}^{2} - β_{n} (1 - β_{n}) {∥ T^{n} y_{n} - x_{n} ∥}^{2} \\ \leq & β_{n} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) (k_{n}^{2} {∥ x_{n} - p ∥}^{2} + α_{n} k_{n}^{2} M) - β_{n} (1 - β_{n}) {∥ T^{n} y_{n} - x_{n} ∥}^{2} \\ = & β_{n} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) k_{n}^{2} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) k_{n}^{2} α_{n} M \\ - β_{n} (1 - β_{n}) {∥ T^{n} y_{n} - x_{n} ∥}^{2} \\ = & (k_{n}^{2} - β_{n} (k_{n}^{2} - 1)) {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) k_{n}^{2} α_{n} M - β_{n} (1 - β_{n}) {∥ T^{n} y_{n} - x_{n} ∥}^{2} . \end{array}

Hence, we have

\begin{aligned} β_{n} (1 - β_{n}) {∥ T^{n} y_{n} - x_{n} ∥}^{2} \\ \leq (k_{n}^{2} - β_{n} (k_{n}^{2} - 1)) {∥ x_{n} - p ∥}^{2} - {∥ x_{n + 1} - p ∥}^{2} + (1 - β_{n}) k_{n}^{2} α_{n} M . \end{aligned}

(3.5)

By the conditions (i), (iii) and

{lim}_{n \to \infty} k_{n} = 1

, we can conclude that

lim_{n \to \infty} ∥ T^{n} y_{n} - x_{n} ∥ = 0 .

(3.6)

Consider that since

y_{n} = P_{C} (x_{n} - λ_{n} \nabla f_{α_{n}} x_{n})

and by Proposition 2.8(ii), we have

\begin{array}{rcl} {∥ y_{n} - p ∥}^{2} & \leq & {∥ x_{n} - λ_{n} \nabla f_{α_{n}} (x_{n}) - p ∥}^{2} - {∥ x_{n} - λ_{n} \nabla f_{α_{n}} (x_{n}) - y_{n} ∥}^{2} \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} 〈 \nabla f_{α_{n}} (x_{n}), p - y_{n} 〉 \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} (〈 \nabla f_{α_{n}} (x_{n}) - \nabla f_{α_{n}} (p), p - x_{n} 〉 \\ + 〈 \nabla f_{α_{n}} (p), p - x_{n} 〉 + 〈 \nabla f_{α_{n}} (x_{n}), x_{n} - y_{n} 〉) \\ \leq & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} (〈 \nabla f_{α_{n}} (p), p - x_{n} 〉 + 〈 \nabla f_{α_{n}} (x_{n}), x_{n} - y_{n} 〉) \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} [〈 (α_{n} I + \nabla f) p, p - x_{n} 〉 + 〈 \nabla f_{α_{n}} (x_{n}), x_{n} - y_{n} 〉] \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} [α_{n} 〈 p, p - x_{n} 〉 + 〈 \nabla f_{α_{n}} (x_{n}), x_{n} - y_{n} 〉] \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} α_{n} 〈 p, p - x_{n} 〉 + 2 λ_{n} 〈 \nabla f_{α_{n}} (x_{n}), x_{n} - y_{n} 〉 \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} α_{n} 〈 p, p - x_{n} 〉 - 2 λ_{n} 〈 \nabla f_{α_{n}} (x_{n}), y_{n} - x_{n} 〉 \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} α_{n} 〈 p, p - x_{n} 〉 - 2 λ_{n} 〈 \nabla f_{α_{n}} (x_{n}), y_{n} - p + p - x_{n} 〉 \\ = & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} α_{n} 〈 p, p - x_{n} 〉 - 2 λ_{n} 〈 \nabla f_{α_{n}} (x_{n}), y_{n} - p 〉 \\ - 2 λ_{n} 〈 \nabla f_{α_{n}} (x_{n}), p - x_{n} 〉 \\ \leq & {∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} α_{n} ∥ p ∥ ∥ p - x_{n} ∥ - 2 λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ y_{n} - p ∥ \\ - 2 λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ p - x_{n} ∥ . \end{array}

(3.7)

Consequently, utilizing Lemma 2.9(ii) and (3.7), we conclude that

\begin{array}{rcl} {∥ x_{n + 1} - p ∥}^{2} & = & {∥ β_{n} x_{n} + (1 - β_{n}) T^{n} y_{n} - p ∥}^{2} \\ = & {∥ β_{n} x_{n} + (1 - β_{n}) T^{n} y_{n} - (β_{n} + (1 - β_{n})) p ∥}^{2} \\ = & {∥ β_{n} x_{n} + (1 - β_{n}) T^{n} y_{n} - β_{n} p - (1 - β_{n}) p ∥}^{2} \\ = & {∥ β_{n} (x_{n} - p) + (1 - β_{n}) (T^{n} y_{n} - p) ∥}^{2} \\ = & β_{n} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) {∥ T^{n} y_{n} - p ∥}^{2} - β_{n} (1 - β_{n}) {∥ x_{n} - T^{n} y_{n} ∥}^{2} \\ \leq & β_{n} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) k_{n}^{2} {∥ y_{n} - p ∥}^{2} - β_{n} (1 - β_{n}) {∥ x_{n} - T^{n} y_{n} ∥}^{2} \\ = & β_{n} {∥ x_{n} - p ∥}^{2} + (1 - β_{n}) k_{n}^{2} [{∥ x_{n} - p ∥}^{2} - {∥ x_{n} - y_{n} ∥}^{2} + 2 λ_{n} α_{n} ∥ p ∥ ∥ p - x_{n} ∥ \\ - 2 λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ y_{n} - p ∥ - 2 λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ p - x_{n} ∥] \\ - β_{n} (1 - β_{n}) {∥ x_{n} - T^{n} y_{n} ∥}^{2} \\ = & (β_{n} + (1 - β_{n}) k_{n}^{2}) {∥ x_{n} - p ∥}^{2} - (1 - β_{n}) k_{n}^{2} {∥ x_{n} - y_{n} ∥}^{2} \\ + 2 (1 - β_{n}) k_{n}^{2} λ_{n} α_{n} ∥ p ∥ ∥ p - x_{n} ∥ \\ - 2 (1 - β_{n}) k_{n}^{2} λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ y_{n} - p ∥ \\ - 2 (1 - β_{n}) k_{n}^{2} λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ p - x_{n} ∥ \\ - β_{n} (1 - β_{n}) {∥ x_{n} - T^{n} y_{n} ∥}^{2} \\ = & (k_{n}^{2} - β_{n} (k_{n}^{2} - 1)) {∥ x_{n} - p ∥}^{2} - (1 - β_{n}) k_{n}^{2} {∥ x_{n} - y_{n} ∥}^{2} \\ + 2 (1 - β_{n}) k_{n}^{2} λ_{n} α_{n} ∥ p ∥ ∥ p - x_{n} ∥ \\ - 2 (1 - β_{n}) k_{n}^{2} λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ y_{n} - p ∥ \\ - 2 (1 - β_{n}) k_{n}^{2} λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ ∥ p - x_{n} ∥ \\ - β_{n} (1 - β_{n}) {∥ x_{n} - T^{n} y_{n} ∥}^{2} . \end{array}

It follows that we get

\begin{aligned} (1 - β_{n}) k_{n}^{2} {∥ x_{n} - y_{n} ∥}^{2} - 2 (1 - β_{n}) k_{n}^{2} λ_{n} α_{n} ∥ p ∥ ∥ p - x_{n} ∥ \\ + 2 (1 - β_{n}) k_{n}^{2} λ_{n} ∥ \nabla f_{α_{n}} (x_{n}) ∥ (∥ y_{n} - p ∥ + ∥ p - x_{n} ∥) + β_{n} (1 - β_{n}) {∥ x_{n} - T^{n} y_{n} ∥}^{2} \\ \leq (k_{n}^{2} - β_{n} (k_{n}^{2} - 1)) {∥ x_{n} - p ∥}^{2} - {∥ x_{n + 1} - p ∥}^{2} . \end{aligned}

(3.8)

So, taking

n \to \infty

, since

{lim}_{n \to 0} k_{n} = 1

, (i)-(iii), (3.6) and (3.8), we can conclude that

lim_{n \to 0} ∥ y_{n} - x_{n} ∥ = 0 .

(3.9)

Consider

\begin{array}{rcl} ∥ x_{n + 1} - x_{n} ∥ & = & ∥ β_{n} x_{n} - x_{n} + (1 - β_{n}) T^{n} y_{n} ∥ \\ = & ∥ - (1 - β_{n}) x_{n} + (1 - β_{n}) T^{n} y_{n} ∥ \\ \leq & (1 - β_{n}) ∥ T^{n} y_{n} - x_{n} ∥ . \end{array}

(3.10)

From (3.6) we obtain

∥ x_{n + 1} - x_{n} ∥ \leq (1 - β_{n}) ∥ T^{n} y_{n} - x_{n} ∥ \to 0 (as n \to \infty) .

(3.11)

Observe that

\begin{array}{rcl} ∥ T^{n} y_{n} - y_{n} ∥ & = & ∥ T^{n} y_{n} - x_{n} + x_{n} - y_{n} ∥ \\ \leq & ∥ T^{n} y_{n} - x_{n} ∥ + ∥ x_{n} - y_{n} ∥ . \end{array}

So, from (3.6) and (3.9), we get

lim_{n \to \infty} ∥ T^{n} y_{n} - y_{n} ∥ = 0 .

(3.12)

We compute that

\begin{array}{rcl} ∥ y_{n + 1} - y_{n} ∥ & = & ∥ P_{C} (x_{n + 1} - λ_{n + 1} \nabla f_{α_{n + 1}} x_{n + 1}) - P_{C} (x_{n} - λ_{n} \nabla f_{α_{n}} x_{n}) ∥ \\ = & ∥ P_{C} (I - λ_{n + 1} \nabla f_{α_{n + 1}}) x_{n + 1} - P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n} ∥ \\ \leq & ∥ P_{C} (I - λ_{n + 1} \nabla f_{α_{n + 1}}) x_{n + 1} - P_{C} (I - λ_{n + 1} \nabla f_{α_{n + 1}}) x_{n} ∥ \\ + ∥ P_{C} (I - λ_{n + 1} \nabla f_{α_{n + 1}}) x_{n} - P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n} ∥ \\ \leq & ∥ x_{n + 1} - x_{n} ∥ + ∥ (I - λ_{n + 1} \nabla f_{α_{n + 1}}) x_{n} - (I - λ_{n} \nabla f_{α_{n}}) x_{n} ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ x_{n} - λ_{n + 1} \nabla f_{α_{n + 1}} x_{n} - (x_{n} - λ_{n} \nabla f_{α_{n}} x_{n}) ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ λ_{n} \nabla f_{α_{n}} x_{n} - λ_{n + 1} \nabla f_{α_{n + 1}} x_{n} ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ λ_{n} (\nabla f + α_{n}) x_{n} - λ_{n + 1} (\nabla f + α_{n + 1}) x_{n} ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ λ_{n} \nabla f x_{n} + λ_{n} α_{n} x_{n} - (λ_{n + 1} \nabla f x_{n} + λ_{n + 1} α_{n + 1} x_{n}) ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ (λ_{n} - λ_{n + 1}) \nabla f x_{n} + λ_{n} α_{n} x_{n} - λ_{n + 1} α_{n + 1} x_{n} ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ (λ_{n} - λ_{n + 1}) \nabla f x_{n} + λ_{n} α_{n} x_{n} - λ_{n} α_{n + 1} x_{n} \\ + λ_{n} α_{n + 1} x_{n} - λ_{n + 1} α_{n + 1} x_{n} ∥ \\ = & ∥ x_{n + 1} - x_{n} ∥ + ∥ (λ_{n} - λ_{n + 1}) \nabla f x_{n} + λ_{n} (α_{n} - α_{n + 1}) x_{n} \\ + (λ_{n} - λ_{n + 1}) α_{n + 1} x_{n} ∥ \\ \leq & ∥ x_{n + 1} - x_{n} ∥ + | λ_{n} - λ_{n + 1} | ∥ \nabla f x_{n} ∥ + λ_{n} | α_{n} - α_{n + 1} | ∥ x_{n} ∥ \\ + α_{n + 1} | λ_{n} - λ_{n + 1} | ∥ x_{n} ∥ . \end{array}

From the conditions (i), (ii) and (3.11), we obtain that

∥ y_{n + 1} - y_{n} ∥ \to 0 (as n \to \infty) .

(3.13)

Since T is uniformly L-Lipschitzian continuous, then

\begin{array}{rcl} ∥ y_{n} - T y_{n} ∥ & \leq & ∥ y_{n} - y_{n + 1} ∥ + ∥ y_{n + 1} - T^{n + 1} y_{n + 1} ∥ + ∥ T^{n + 1} y_{n + 1} - T^{n + 1} y_{n} ∥ + ∥ T^{n + 1} y_{n} - T y_{n} ∥ \\ \leq & ∥ y_{n} - y_{n + 1} ∥ + ∥ y_{n + 1} - T^{n + 1} y_{n + 1} ∥ + L ∥ y_{n} - y_{n + 1} ∥ + L ∥ T^{n} y_{n} - y_{n} ∥ . \end{array}

Since

{lim}_{n \to \infty} ∥ y_{n + 1} - y_{n} ∥ = 0

and

{lim}_{n \to \infty} ∥ y_{n} - T^{n} y_{n} ∥ = 0

, it follows that

lim_{n \to \infty} ∥ y_{n} - T y_{n} ∥ = 0 .

(3.14)

Step 3. We show that

\hat{x} \in Fix (T) \cap Γ

Since

\nabla f = A^{*} (I - P_{Q}) A

is Lipschitz continuous, from (3.9), we have

lim_{n \to \infty} ∥ \nabla f (x_{n}) - \nabla f (y_{n}) ∥ = 0 .

Since

{x_{n}}

is bounded, there is a subsequence

{x_{n_{i}}}

{x_{n}}

that converges weakly to some

\hat{x}

First, we show that

\hat{x} \in Γ

. Since

∥ x_{n} - y_{n} ∥ \to 0

, it is known that

y_{n_{i}} ⇀ \hat{x}

Put

S w_{1} = {\begin{matrix} \nabla f w_{1} + N_{C} w_{1} & if w_{1} \in C, \\ \emptyset & if w_{1} \notin C, \end{matrix}

where

N_{C} w_{1} = {z \in H_{1} : 〈 w_{1} - u, z 〉 \geq 0, \forall u \in C}

. Then S is maximal monotone and

0 \in S w_{1}

if and only if

w_{1} \in VI (C, \nabla f)

; (see [17]) for more details. Let

(w_{1}, z) \in G (S)

, we have

z \in S w_{1} = \nabla f w_{1} + N_{C} w_{1},

and hence

z - \nabla f w_{1} \in N_{C} w_{1} .

So, we have

〈 w_{1} - u, z - \nabla f w_{1} 〉 \geq 0, \forall u \in C .

On the other hand, from

y_{n} = P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n} and w_{1} \in C,

we have

〈 x_{n} - λ_{n} \nabla f_{α_{n}} x_{n} - y_{n}, y_{n} - w_{1} 〉 \geq 0,

and

〈 w_{1} - y_{n}, \frac{y_{n} - x_{n}}{λ_{n}} + \nabla f_{α_{n}} x_{n} 〉 \geq 0 .

Therefore, from

z - \nabla f w_{1} \in N_{C} w_{1}

and

y_{n_{i}} \in C

, it follows that

\begin{array}{rcl} 〈 w_{1} - y_{n_{i}}, z 〉 & \geq & 〈 w_{1} - y_{n_{i}}, \nabla f w_{1} 〉 \\ \geq & 〈 w_{1} - y_{n_{i}}, \nabla f w_{1} 〉 - 〈 w_{1} - y_{n_{i}}, \frac{y_{n_{i}} - x_{n_{i}}}{λ_{n_{i}}} + \nabla f_{α_{n_{i}}} x_{n_{i}} 〉 \\ = & 〈 w_{1} - y_{n_{i}}, \nabla f w_{1} 〉 - 〈 w_{1} - y_{n_{i}}, \frac{y_{n_{i}} - x_{n_{i}}}{λ_{n_{i}}} + \nabla f x_{n_{i}} 〉 \\ - α_{n_{i}} 〈 w_{1} - y_{n_{i}}, x_{n_{i}} 〉 \\ = & 〈 w_{1} - y_{n_{i}}, \nabla f w_{1} - \nabla f y_{n_{i}} 〉 + 〈 w_{1} - y_{n_{i}}, \nabla f y_{n_{i}} - \nabla f x_{n_{i}} 〉 \\ - 〈 w_{1} - y_{n_{i}}, \frac{y_{n_{i}} - x_{n_{i}}}{λ_{n_{i}}} 〉 - α_{n_{i}} 〈 w_{1} - y_{n_{i}}, x_{n_{i}} 〉 \\ \geq & 〈 w_{1} - y_{n_{i}}, \nabla f y_{n_{i}} - \nabla f x_{n_{i}} 〉 - 〈 w_{1} - y_{n_{i}}, \frac{y_{n_{i}} - x_{n_{i}}}{λ_{n_{i}}} 〉 \\ - α_{n_{i}} 〈 w_{1} - y_{n_{i}}, x_{n_{i}} 〉 . \end{array}

Hence, we obtain

〈 w_{1} - \hat{x}, z 〉 \geq 0 as i \to \infty .

Since S is maximal monotone, we have

\hat{x} \in S^{- 1} 0

, and hence

\hat{x} \in VI (C, \nabla f)

. Thus, it is clear that

\hat{x} \in Γ

Next, we show that

\hat{x} \in Fix (T)

. Indeed, since

y_{n_{i}} ⇀ \hat{x}

and

∥ y_{n_{i}} - T y_{n_{i}} ∥ \to 0

by (3.14) and Lemma 2.6, we get

\hat{x} \in Fix (T)

. Therefore, we have

\hat{x} \in Fix (T) \cap Γ

Let

{x_{n_{j}}}

be another subsequence of

{x_{n}}

such that

{x_{n_{j}}} ⇀ \bar{x}

. Then

\bar{x} \in Fix (T) \cap Γ

. Let us show that

\hat{x} = \bar{x}

. Assume that

\hat{x} \neq \bar{x}

. From the Opial condition [18], we have

\begin{array}{rcl} lim_{n \to \infty} ∥ x_{n} - \hat{x} ∥ & = & lim_{n_{i} \to \infty} inf ∥ x_{n_{i}} - \hat{x} ∥ \\ < & lim_{n_{i} \to \infty} inf ∥ x_{n_{i}} - \bar{x} ∥ \\ = & lim_{n \to \infty} ∥ x_{n} - \bar{x} ∥ \\ = & lim_{n_{j} \to \infty} inf ∥ x_{n_{j}} - \bar{x} ∥ \\ < & lim_{n_{j} \to \infty} inf ∥ x_{n_{j}} - \hat{x} ∥ \\ = & lim_{n \to \infty} ∥ x_{n} - \hat{x} ∥ . \end{array}

This is a contradiction. Thus, we have

\hat{x} = \bar{x}

. This implies

x_{n} ⇀ \hat{x} \in Fix (T) \cap Γ .

Further, from

∥ x_{n} - y_{n} ∥ \to 0

, it follows that

y_{n} ⇀ \hat{x}

. This shows that both sequences

{x_{n}}

and

{y_{n}}

converge weakly to

\hat{x} \in Fix (T) \cap Γ

. This completes the proof. □

Utilizing Theorem 3.1, we have the following new results in the setting of real Hilbert spaces.

Take

T^{n} \equiv I (identity mappings)

in Theorem 3.1. Therefore the conclusion follows.

Corollary 3.2LetCbe a nonempty, closed, and convex subset of a real Hilbert spaceH. Suppose that

Γ \neq \emptyset

. Let

{x_{n}}

be a sequence inCgenerated by the following algorithm:

{\begin{matrix} x_{0} = x \in C chosen arbitrarily, \\ x_{n + 1} = β_{n} x_{n} + (1 - β_{n}) P_{C} (I - λ_{n} \nabla f_{α_{n}}) x_{n}, \forall n \geq 0, \end{matrix}

(3.15)

where

\nabla f_{α_{n}} = \nabla f + α_{n} I = A^{*} (I - P_{Q}) A + α_{n} I

, and the sequences

{α_{n}}

{λ_{n}}

, and

{β_{n}}

satisfy the conditions:

(i)

\sum_{n = 1}^{\infty} α_{n} < \infty

(ii)

{λ_{n}} \subset [a, b]

for some

a, b \in (0, \frac{1}{{∥ A ∥}^{2}})

and

\sum_{n = 1}^{\infty} | λ_{n + 1} - λ_{n} | < \infty

(iii)

{β_{n}} \subset [c, d]

for some

c, d \in (0, 1)

Then the sequence

{x_{n}}

converges weakly to an element

\hat{x} \in Γ

Take

P_{C} \equiv I (identity mappings)

in Theorem 3.1. Therefore the conclusion follows.

Corollary 3.3LetCbe a nonempty, closed, and convex subset of a real Hilbert spaceHand let

T : C \to C

be a uniformlyL-Lipschitzian and quasi-nonexpansive mapping with

Fix (T) \neq \emptyset

and

{k_{n}} \subset [1, \infty)

for all

n \in N

such that

\sum_{n = 1}^{\infty} (k_{n} - 1) < \infty

. Let

{x_{n}}

be the sequence inCgenerated by the following algorithm:

{\begin{matrix} x_{0} = x \in C chosen arbitrarily, \\ x_{n + 1} = β_{n} x_{n} + (1 - β_{n}) T^{n} x_{n}, \forall n \geq 0, \end{matrix}

(3.16)

and let the sequence

{β_{n}}

satisfy the condition

{β_{n}} \subset [c, d]

for some

c, d \in (0, 1)

. Then the sequence

{x_{n}}

converges weakly to an element

\hat{x} \in Fix (T)

Remark 3.4 Theorem 3.1 improves and extends [[7], Theorem 5.7] in the following aspects:

(a)

The iterative algorithm [[7], Theorem 5.7] is extended for developing our relaxed extragradient algorithm with regularization in Theorem 3.1.

(b)

The technique of proving weak convergence in Theorem 3.1 is different from that in [[7], Theorem 5.7] because of our technique to use asymptotically quasi-nonexpansive mappings and the property of maximal monotone mappings.

(c)

The problem of finding a common element of

Fix (T) \cap Γ

for asymptotically quasi-nonexpansive mappings which is more general than that for nonexpansive mappings and the problem of finding a solution of the SFP in [[7], Theorem 5.7].

Acknowledgements

The authors thank the referee for comments and suggestions on this manuscript. The first author was supported by the Thailand Research Fund through the Royal Golden Jubilee Ph.D. Program (Grant No. PHD/0033/2554) and the King Mongkut’s University of Technology Thonburi (KMUTT). Moreover, this study was supported by the Higher Education Research Promotion and National Research University Project of Thailand, Office of the Higher Education Commission (Under Grant No. NRU56000508).

Open AccessThis 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.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

All authors contributed equally and significantly in writing this paper. All authors read and approved the final manuscript.

Vorheriger Artikel Some improvements of Minkowski’s integral inequality on time scales

Nächster Artikel Asymptotic decay estimate of solutions to the generalized damped Bq equation

Censor Y, Elving T: A multiprojection algorithm using Bregman projections in product space. Numer. Algorithms 1994, 8: 221–239. 10.1007/BF02142692MathSciNetCrossRefMATH

Byrne C: Iterative oblique projection onto convex subsets and the split feasibility problem. Inverse Probl. 2002, 18: 441–453. 10.1088/0266-5611/18/2/310MathSciNetCrossRefMATH

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. 10.1088/0031-9155/51/10/001CrossRef

Censor Y, Elving T, Kopf N, Bortfeld T: The multiple-set split feasibility problem and its applications for inverse problem. Inverse Probl. 2005, 21: 2071–2084. 10.1088/0266-5611/21/6/017CrossRefMATH

Censor Y, Motova A, Segal A: Perturbed projections and subgradient projections for the multiple-set split feasibility problem. J. Math. Anal. Appl. 2007, 327: 1244–1256. 10.1016/j.jmaa.2006.05.010MathSciNetCrossRefMATH

Censor Y, Segal A: Iterative projection methods in biomedical inverse problem. In Mathematical Methods in Biomedical Imaging and Intensity-Modulate Therapy, IMRT. Edited by: Censor Y, Jiang M, Louis AK. Edizioni della Normale, Pisa; 2008:65–96.

Xu HK: Iterative methods for the split feasibility problem in infinite-dimensional Hilbert spaces. Inverse Probl. 2010., 26(10): Article ID 105018

Landweber L: An iterative formula for Fredholm integral equations of the first kind. Am. J. Math. 1951, 73: 615–625. 10.2307/2372313MathSciNetCrossRefMATH

Byrne C: A unified treatment of some iterative algorithms in signal processing and image reconstruction. Inverse Probl. 2004, 26: 103–120.MathSciNetCrossRefMATH

10.

Combettes PL, Wajs V: Signal recovery by proximal forward-backward splitting multiscale model. Simulation 2005, 4: 1168–1200.MathSciNetMATH

11.

Deepho J, Kumam P: A modified Halpern’s iterative scheme for solving split feasibility problems. Abstr. Appl. Anal. 2012., 2012: Article ID 876069

12.

Ceng LC, Ansari QH, Yao JC: Relaxed extragradient method for finding minimum-norm solutions of the split feasibility problems. Nonlinear Anal. 2012, 75(4):2116–2125. 10.1016/j.na.2011.10.012MathSciNetCrossRefMATH

13.

Combettes PL: Solving monotone inclusions via compositions of nonexpansive averaged operator. Optimization 2004, 53(5–6):475–504. 10.1080/02331930412331327157MathSciNetCrossRefMATH

14.

Geobel K, Kirk WA Cambridge Studies in Advanced Mathematics 28. In Topics in Metric Fixed Point Theory. Cambridge University Press, Cambridge; 1990.CrossRef

15.

Tan KK, Xu HK: Approximating fixed points of nonexpansive mappings by the Ishikawa iteration process. J. Math. Anal. Appl. 1993, 178: 301–308. 10.1006/jmaa.1993.1309MathSciNetCrossRefMATH

16.

Marino G, Xu HK: Weak and strong convergence theorems for strict pseudo-contractions in Hilbert space. J. Math. Anal. Appl. 2007, 329: 336–346. 10.1016/j.jmaa.2006.06.055MathSciNetCrossRefMATH

17.

Rockafellar RT: On the maximality of sums of nonlinear monotone operators. Trans. Am. Math. Soc. 1970, 149: 75–88. 10.1090/S0002-9947-1970-0282272-5MathSciNetCrossRefMATH

18.

Opial Z: Weak convergence of the sequence of successive approximations for nonexpansive mappings. Bull. Am. Math. Soc. 1967, 73: 591–597. 10.1090/S0002-9904-1967-11761-0MathSciNetCrossRefMATH

Titel: Split feasibility and fixed-point problems for asymptotically quasi-nonexpansive mappings
verfasst von: Jitsupa Deepho
Poom Kumam
Publikationsdatum: 01.12.2013
Verlag: Springer International Publishing
Erschienen in: Journal of Inequalities and Applications / Ausgabe 1/2013
Elektronische ISSN: 1029-242X
DOI: https://doi.org/10.1186/1029-242X-2013-322

Springer Professional

Split feasibility and fixed-point problems for asymptotically quasi-nonexpansive mappings

Abstract

Competing interests

Authors’ contributions

1 Introduction

2 Preliminaries

3 Main result

Acknowledgements

Competing interests

Authors’ contributions

Premium Partner

Springer Professional

Abstract

Competing interests

Authors’ contributions

1 Introduction

2 Preliminaries

3 Main result

Acknowledgements

Competing interests

Authors’ contributions

Weitere Artikel der Ausgabe 1/2013

The general iterative methods for equilibrium problems and fixed point problems of a countable family of nonexpansive mappings in Hilbert spaces

Hausdorff measure of noncompactness in some sequence spaces of a triple band matrix

Convergence of Halley’s method for operators with the bounded second Fréchet-derivative in Banach spaces

Retracted: Coupled fixed point theorems without continuity and mixed monotone property

Applications of Kato’s inequality for n-tuples of operators in Hilbert spaces, (I)

Statistical inference for the shape parameter change-point estimator in negative associated gamma distribution

Premium Partner