 Research
 Open Access
An implicitly defined iterative sequence for monotone operators in Banach spaces
 Fumiaki Kohsaka^{1}Email author
https://doi.org/10.1186/1029242X2014181
© Kohsaka; licensee Springer. 2014
 Received: 31 October 2013
 Accepted: 22 April 2014
 Published: 12 May 2014
Abstract
Given a monotone operator in a Banach space, we show that an iterative sequence, which is implicitly defined by a fixed point theorem for mappings of firmly nonexpansive type, converges strongly to a minimum norm zero point of the given operator. Applications to a convex minimization problem and a variational inequality problem are also included.
MSC:47H05, 47J25, 47N10.
Keywords
 Banach space
 convex minimization
 mapping of firmly nonexpansive type
 fixed point
 monotone operator
 proximal point method
 variational inequality
1 Introduction
Many nonlinear problems can be formulated as the problem of finding a zero point of a maximal monotone operator in a Banach space. The proximal point method, which was first introduced by Martinet [1] and generally studied by Rockafellar [2], is an iterative method for approximating a solution to this problem.
for all $n\in \mathbb{N}$, where $A:X\to {2}^{{X}^{\ast}}$ is a maximal monotone operator, X is a smooth, strictly convex, and reflexive real Banach space, $J:X\to {X}^{\ast}$ is the normalized duality mapping, and $\{{\lambda}_{n}\}$ is a sequence of positive real numbers.
The following result was obtained in [3]: If ${\sum}_{n=1}^{\mathrm{\infty}}{\lambda}_{n}=\mathrm{\infty}$, then $\{{x}_{n}\}$ is bounded if and only if ${A}^{1}0$ is nonempty. Further, if X is uniformly convex, the norm of X is uniformly Gâteaux differentiable, ${A}^{1}0$ is nonempty, ${inf}_{n}{\lambda}_{n}>0$, and J is weakly sequentially continuous, then $\{{x}_{n}\}$ converges weakly to an element of ${A}^{1}0$. This is a generalization of the result due to Rockafellar [2] in Hilbert spaces. See also [4, 5] for some related results.
for all $n\in \mathbb{N}$, where A, X, J, and $\{{\lambda}_{n}\}$ are the same as in (1.1) and $\{{\beta}_{n}\}$ is a sequence of $[0,1)$. Under some additional assumptions, we show that $\{{x}_{n}\}$ is well defined and is strongly convergent to an element of ${A}^{1}0$ of minimal norm; see Theorem 3.5.
The schemes (1.1) and (1.2) above are similar to each other, though their properties are quite different. In fact, the former fails to converge strongly even in Hilbert spaces [6], whereas the latter converges strongly in Banach spaces. Further, the former is well defined since $\mathrm{R}(J+\lambda A)={X}^{\ast}$ for each $\lambda >0$, whereas the latter is not necessarily well defined. To study the well definedness and the asymptotic behavior of $\{{x}_{n}\}$ in (1.2), we exploit some techniques in [3, 7].
This paper is organized as follows: In Section 2, we give some definitions, recall some known results, and briefly study the existence of a zero point of a monotone operator. In Section 3, using the results in the previous section, we first obtain a convergence theorem for a monotone operator satisfying a range condition; see Theorem 3.1. Using this result, we show a convergence theorem for a maximal monotone operator; see Theorem 3.5. In Section 4, we apply Theorem 3.5 to a convex minimization problem and a variational inequality problem.
2 Preliminaries
Throughout the present paper, we denote by ℕ the set of all positive integers, ℝ the set of all real numbers, X a smooth, strictly convex, and reflexive real Banach space with dual ${X}^{\ast}$, $\parallel \cdot \parallel $ the norms of X and ${X}^{\ast}$, $\u3008x,{x}^{\ast}\u3009$ the value of ${x}^{\ast}\in {X}^{\ast}$ at $x\in X$, ${x}_{n}\to x$ the strong convergence of a sequence $\{{x}_{n}\}$ of X to $x\in X$, ${x}_{n}\rightharpoonup x$ the weak convergence of a sequence $\{{x}_{n}\}$ of X to $x\in X$, $\overline{U}$ the norm closure of $U\subset X$, coU the convex hull of $U\subset X$, $\overline{co}\phantom{\rule{0.2em}{0ex}}U$ the closed convex hull of $U\subset X$, and ${S}_{X}$ the unit sphere of X, respectively.
converges uniformly in $x\in {S}_{X}$ for all $y\in {S}_{X}$. The space X is said to be uniformly convex if for each $\epsilon \in (0,2]$, there exists $\delta >0$ such that $\parallel (x+y)/2\parallel \le 1\delta $ whenever $x,y\in {S}_{X}$ and $\parallel xy\parallel \ge \epsilon $. The space X is said to have the KadecKlee property if ${x}_{n}\to x$ whenever $\{{x}_{n}\}$ is a sequence of X such that ${x}_{n}\rightharpoonup x\in X$ and $\parallel {x}_{n}\parallel \to \parallel x\parallel $. Every uniformly convex Banach space is both strictly convex and reflexive and has the KadecKlee property; see [8, 9].
for all $x,y\in X$.
for all $x,y\in C$; see also [13]. If X is a Hilbert space, then $T:C\to X$ is firmly nonexpansive if and only if it is of firmly nonexpansive type.
For an operator $A:X\to {2}^{{X}^{\ast}}$, the domain $\mathrm{D}(A)$, the range $\mathrm{R}(A)$, and the graph $\mathrm{G}(A)$ of A are defined by $\mathrm{D}(A)=\{x\in X:Ax\ne \mathrm{\varnothing}\}$, $\mathrm{R}(A)={\bigcup}_{x\in X}Ax$, and $\mathrm{G}(A)=\{(x,{x}^{\ast})\in X\times {X}^{\ast}:{x}^{\ast}\in Ax\}$, respectively. The operator A is said to be monotone if $\u3008xy,{x}^{\ast}{y}^{\ast}\u3009\ge 0$ whenever $(x,{x}^{\ast}),(y,{y}^{\ast})\in \mathrm{G}(A)$. It is also said to be maximal monotone if A is monotone and there is no monotone operator $B:X\to {2}^{{X}^{\ast}}$ such that $A\ne B$ and $\mathrm{G}(A)\subset \mathrm{G}(B)$. Let C be a nonempty closed convex subset of X and $A:X\to {2}^{{X}^{\ast}}$ a monotone operator such that $\mathrm{D}(A)\subset C\subset {J}^{1}\mathrm{R}(J+A)$. Then the mapping $T:C\to C$ defined by $Tx={(J+A)}^{1}Jx$ for all $x\in C$ is of firmly nonexpansive type and $\mathrm{F}(T)={A}^{1}0$; see [12, 14]. We know the following lemma.
Lemma 2.1 ([3])
$\{{\lambda}_{n}\}$ a sequence of $(0,\mathrm{\infty})$ such that ${inf}_{n}{\lambda}_{n}>0$, and ${Q}_{{\lambda}_{n}}:C\to C$ the mapping defined by ${Q}_{{\lambda}_{n}}x={(J+{\lambda}_{n}A)}^{1}Jx$ for all $x\in C$ and $n\in \mathbb{N}$. If $\{{x}_{n}\}$ is a sequence of C such that ${x}_{n}\rightharpoonup u$ and ${x}_{n}{Q}_{{\lambda}_{n}}{x}_{n}\to 0$, then u is an element of ${A}^{1}0$.
We know the following result for mappings of firmly nonexpansive type.
Lemma 2.2 ([12])
 (i)
$\mathrm{F}(T)$ is nonempty if and only if $\{{T}^{n}x\}$ is bounded for some $x\in C$;
 (ii)
$\mathrm{F}(T)$ is closed and convex.
Using Lemma 2.2, we can show the following.
Lemma 2.3 Let C be a nonempty closed convex subset of X and $T:C\to C$ a mapping of firmly nonexpansive type. Suppose that $\beta \in [0,1)$, $0\in C$, and $T(C)$ is bounded. Then the mapping βT has a unique fixed point.
for all $x,y\in C$. This implies that S is of firmly nonexpansive type. Since C is convex, $0\in C$, and $\beta \in [0,1)$, we know that S is a mapping of C into itself. Further, since $S(C)=\beta T(C)$ and $T(C)$ is bounded, the sequence $\{{S}^{n}x\}$ is bounded for all $x\in C$. Thus Lemma 2.2 implies that $\mathrm{F}(S)$ is nonempty.
We next show that $\mathrm{F}(S)$ consists of one point. Suppose that $p,{p}^{\prime}\in \mathrm{F}(S)$. Then it follows from (2.8) that $(1\beta )\u3008p{p}^{\prime},JpJ{p}^{\prime}\u3009=0$. Since $1\beta >0$, we obtain $\u3008p{p}^{\prime},JpJ{p}^{\prime}\u3009=0$. Thus the strict monotonicity of J implies that $p={p}^{\prime}$. □
As a direct consequence of Lemmas 2.2 and 2.3, we obtain the following.
 (i)
${A}^{1}0$ is nonempty, closed, and convex;
 (ii)if $0\in C$ and $\beta \in [0,1)$, then there exists a unique $p\in C$ such that$p=\beta {(J+A)}^{1}Jp.$(2.9)
Proof Let $T:C\to C$ be the mapping defined by $Tx={(J+A)}^{1}Jx$ for all $x\in C$. Then we know that T is of firmly nonexpansive type and $T(C)\subset \mathrm{D}(A)$. Hence $\{{T}^{n}x\}$ is bounded for all $x\in C$. On the other hand, we know that $\mathrm{F}(T)={A}^{1}0$. Therefore, part (i) follows from Lemma 2.2. Part (ii) follows from Lemma 2.3. □
3 Strong convergence of an iterative sequence
In this section, we first show the following strong convergence theorem for a monotone operator satisfying a range condition.
 (i)
For each $n\in \mathbb{N}$, there exists a unique ${x}_{n}\in C$ such that ${x}_{n}={\beta}_{n}{Q}_{{\lambda}_{n}}{x}_{n}$;
 (ii)
if X has the KadecKlee property, the norm of X is uniformly Gâteaux differentiable, ${inf}_{n}{\lambda}_{n}>0$, and ${lim}_{n}{\beta}_{n}=1$, then the sequence $\{{x}_{n}\}$ converges strongly to ${P}_{{A}^{1}0}(0)$.
Part (i) of Theorem 3.1 follows from Corollary 2.4.
The proof of (i) of Theorem 3.1 Let $n\in \mathbb{N}$ be given and set $B={\lambda}_{n}A$. Then $B:X\to {2}^{{X}^{\ast}}$ is monotone and $\mathrm{D}(B)=\mathrm{D}(A)$. Thus we know that $\mathrm{D}(B)$ is bounded and $\mathrm{D}(B)\subset C\subset {J}^{1}\mathrm{R}(J+B)$. Therefore, part (ii) of Corollary 2.4 ensures the conclusion. □
Before proving (ii) of Theorem 3.1, we show the following lemma.
 (i)
$\u3008{Q}_{{\lambda}_{n}}{x}_{n}y,J{Q}_{{\lambda}_{n}}{x}_{n}\u3009\le 0$ for all $y\in {A}^{1}0$ and $n\in \mathbb{N}$;
 (ii)
$\varphi ({Q}_{{\lambda}_{n}}{x}_{n},y)+\varphi (y,{Q}_{{\lambda}_{n}}{x}_{n})\le 2\u3008y{Q}_{{\lambda}_{n}}{x}_{n},Jy\u3009$ for all $y\in {A}^{1}0$ and $n\in \mathbb{N}$.
By (3.2), (3.3), and $1{\beta}_{n}>0$, the result follows.
for all $y\in {A}^{1}0$ and $n\in \mathbb{N}$. Thus the result follows. □
We next show (ii) of Theorem 3.1.
The proof of (ii) of Theorem 3.1 Suppose that X has the KadecKlee property, the norm of X is uniformly Gâteaux differentiable, ${inf}_{n}{\lambda}_{n}>0$, and ${lim}_{n}{\beta}_{n}=1$. Set ${y}_{n}={Q}_{{\lambda}_{n}}{x}_{n}$ for all $n\in \mathbb{N}$. By (i) of Corollary 2.4, the set ${A}^{1}0$ is nonempty, closed, and convex. Hence ${P}_{{A}^{1}0}$ is well defined. We denote ${P}_{{A}^{1}0}$ by P.
Since ${y}_{n}\in \mathrm{D}(A)$ for all $n\in \mathbb{N}$ and $\mathrm{D}(A)$ is bounded, $\{{y}_{n}\}$ is bounded. By the definition of $\{{x}_{n}\}$, we have $\parallel {x}_{n}\parallel ={\beta}_{n}\parallel {y}_{n}\parallel \le \parallel {y}_{n}\parallel $ for all $n\in \mathbb{N}$. Thus $\{{x}_{n}\}$ is also bounded.
Since ${inf}_{j}{\lambda}_{{n}_{{i}_{j}}}\ge {inf}_{n}{\lambda}_{n}>0$, Lemma 2.1 shows that u is an element of ${A}^{1}0$. It also follows from (3.5) and ${x}_{{n}_{{i}_{j}}}\rightharpoonup u$ that ${y}_{{n}_{{i}_{j}}}\rightharpoonup u$.
Thus we obtain $\parallel {y}_{{n}_{{i}_{j}}}\parallel \to \parallel u\parallel $. Since X has the KadecKlee property, we have ${y}_{{n}_{{i}_{j}}}\to u$. Consequently, it follows from (3.5) that ${x}_{{n}_{{i}_{j}}}\to u$.
Noting that $u\in {A}^{1}0$, we have from (2.2) and (3.8) that $u=P(0)$. Therefore, we conclude that $\{{x}_{n}\}$ converges strongly to $P(0)$. □
As a direct consequence of Theorem 3.1, we obtain the following corollary.
 (i)
For each $n\in \mathbb{N}$, there exists a unique ${x}_{n}\in C$ such that ${x}_{n}={\beta}_{n}T{x}_{n}$;
 (ii)
if X has the KadecKlee property, the norm of X is uniformly Gâteaux differentiable, and ${lim}_{n}{\beta}_{n}=1$, then the sequence $\{{x}_{n}\}$ converges strongly to ${P}_{\mathrm{F}(T)}(0)$.
for all $x\in X$. Then, by [14], we know that the following hold:

A is a monotone operator and ${A}^{1}0=\mathrm{F}(T)$;

$\mathrm{D}(A)=T(C)\subset C={J}^{1}\mathrm{R}(J+A)$;

$Tx={(J+A)}^{1}Jx$ for all $x\in C$.
Thus the result follows from Theorem 3.1. □
Remark 3.4 In the case when C is bounded, Corollary 3.3 is reduced to the result obtained in [7].
Using Theorem 3.1, we can also show the following strong convergence theorem for a maximal monotone operator.
Theorem 3.5 Let X be a smooth, strictly convex, and reflexive real Banach space, $A:X\to {2}^{{X}^{\ast}}$ a maximal monotone operator such that $0\in \overline{\mathrm{D}(A)}$ and $\mathrm{D}(A)$ is bounded, $\{{\lambda}_{n}\}$ a sequence of positive real numbers, and $\{{\beta}_{n}\}$ a sequence of $[0,1)$. Then for each $n\in \mathbb{N}$, there exists a unique ${x}_{n}\in \overline{\mathrm{D}(A)}$ satisfying (1.2). Moreover, if X has the KadecKlee property, the norm of X is uniformly Gâteaux differentiable, ${inf}_{n}{\lambda}_{n}>0$, and ${lim}_{n}{\beta}_{n}=1$, then the sequence $\{{x}_{n}\}$ converges strongly to ${P}_{{A}^{1}0}(0)$.
for all $x\in \overline{co}\phantom{\rule{0.2em}{0ex}}\mathrm{D}(A)$, where I denotes the identity mapping on X; see [16, 17]. Since ${(I+\mu {J}^{1}A)}^{1}x\in \mathrm{D}(A)$ for all $\mu >0$ and $x\in X$, it follows from (3.10) that $\overline{co}\phantom{\rule{0.2em}{0ex}}\mathrm{D}(A)\subset \overline{\mathrm{D}(A)}$. Thus $\overline{co}\phantom{\rule{0.2em}{0ex}}\mathrm{D}(A)=\overline{\mathrm{D}(A)}$ and hence $\overline{\mathrm{D}(A)}$ is convex.
Noting that $0\in \overline{\mathrm{D}(A)}=C$, we obtain the desired result by Theorem 3.1. □
4 Results deduced from Theorem 3.5
In this final section, we study two applications of Theorem 3.5. Throughout this section, we suppose the following:

X is a uniformly convex real Banach space whose norm is uniformly Gâteaux differentiable;

$\{{\lambda}_{n}\}$ is a sequence of positive real numbers such that ${inf}_{n}{\lambda}_{n}>0$;

$\{{\beta}_{n}\}$ is a sequence of $[0,1)$ such that ${lim}_{n}{\beta}_{n}=1$.
We first study a convex minimization problem. For a function $f:X\to (\mathrm{\infty},\mathrm{\infty}]$, we denote by argminf or ${argmin}_{y\in X}f(y)$ the set of all $u\in X$ such that $f(u)=inff(X)$. In the case when $argminf=\{p\}$ for some $p\in X$, we identify argminf with p. The set of all $x\in X$ such that $f(x)\in \mathbb{R}$ is denoted by $\mathrm{D}(f)$. We denote by ∂f the subdifferential mapping of f; see [17, 19] for more details.
Moreover, the sequence $\{{x}_{n}\}$ converges strongly to ${P}_{argminf}(0)$.
holds for all $\lambda >0$ and $x\in X$. Thus we know that ${x}_{n}={\beta}_{n}{(J+{\lambda}_{n}A)}^{1}J{x}_{n}$ for all $n\in \mathbb{N}$. Consequently, Theorem 3.5 implies the conclusion. □
We finally study a variational inequality problem. For a nonempty closed convex subset C of X and an operator $B:C\to {X}^{\ast}$, we denote by $VI(C,B)$ the set of all $u\in C$ such that $\u3008yu,Bu\u3009\ge 0$ for all $y\in C$. In the case when $VI(C,B)=\{p\}$ for some $p\in C$, we identify $VI(C,B)$ with p. The operator B is said to be hemicontinuous if the mapping $g:[0,1]\to {X}^{\ast}$ defined by $g(t)=B(tx+(1t)y)$ for all $t\in [0,1]$ is continuous with respect to the weak* topology in ${X}^{\ast}$ for all $x,y\in C$.
Moreover, the sequence $\{{x}_{n}\}$ converges strongly to ${P}_{VI(C,B)}(0)$.
holds for all $\lambda >0$ and $x\in X$. Thus we have ${x}_{n}={\beta}_{n}{(J+{\lambda}_{n}A)}^{1}J{x}_{n}$ for all $n\in \mathbb{N}$. Consequently, Theorem 3.5 implies the conclusion. □
Declarations
Acknowledgements
The author would like to thank the anonymous referees for carefully reading the original version of the manuscript. The author is supported by GrantinAid for Young Scientists No. 25800094 from the Japan Society for the Promotion of Science.
Authors’ Affiliations
References
 Martinet B: Régularisation d’inéquations variationnelles par approximations successives. Rev. Fr. Inform. Rech. Opér. 1970, 4: 154–158. (in French)MathSciNetGoogle Scholar
 Rockafellar RT: Monotone operators and the proximal point algorithm. SIAM J. Control Optim. 1976, 14: 877–898. 10.1137/0314056MathSciNetView ArticleGoogle Scholar
 Aoyama K, Kohsaka F, Takahashi W: Proximal point methods for monotone operators in Banach spaces. Taiwanese J. Math. 2011, 15: 259–281.MathSciNetGoogle Scholar
 Kamimura S: The proximal point algorithm in a Banach space. In Nonlinear Analysis and Convex Analysis. Yokohama Publishers, Yokohama; 2004:143–148.Google Scholar
 Kamimura S, Kohsaka F, Takahashi W: Weak and strong convergence theorems for maximal monotone operators in a Banach space. SetValued Anal. 2004, 12: 417–429. 10.1007/s1122800481964MathSciNetView ArticleGoogle Scholar
 Güler O: On the convergence of the proximal point algorithm for convex minimization. SIAM J. Control Optim. 1991, 29: 403–419. 10.1137/0329022MathSciNetView ArticleGoogle Scholar
 Kohsaka F, Takahashi W: Strongly convergent net given by a fixed point theorem for firmly nonexpansive type mappings. Appl. Math. Comput. 2008, 202: 760–765. 10.1016/j.amc.2008.03.019MathSciNetView ArticleGoogle Scholar
 Cioranescu I: Geometry of Banach Spaces, Duality Mappings and Nonlinear Problems. Kluwer Academic, Dordrecht; 1990.View ArticleGoogle Scholar
 Takahashi W: Nonlinear Functional Analysis. Yokohama Publishers, Yokohama; 2000.Google Scholar
 Alber YI: Metric and generalized projection operators in Banach spaces: properties and applications. Lecture Notes in Pure and Appl. Math. 178. In Theory and Applications of Nonlinear Operators of Accretive and Monotone Type. Dekker, New York; 1996:15–50.Google Scholar
 Kamimura S, Takahashi W: Strong convergence of a proximaltype algorithm in a Banach space. SIAM J. Optim. 2002, 13: 938–945. 10.1137/S105262340139611XMathSciNetView ArticleGoogle Scholar
 Kohsaka F, Takahashi W: Existence and approximation of fixed points of firmly nonexpansivetype mappings in Banach spaces. SIAM J. Optim. 2008, 19: 824–835. 10.1137/070688717MathSciNetView ArticleGoogle Scholar
 Aoyama K, Kohsaka F, Takahashi W: Three generalizations of firmly nonexpansive mappings: their relations and continuity properties. J. Nonlinear Convex Anal. 2009, 10: 131–147.MathSciNetGoogle Scholar
 Kohsaka F, Takahashi W: Fixed point theorems for a class of nonlinear mappings related to maximal monotone operators in Banach spaces. Arch. Math. 2008, 91: 166–177. 10.1007/s0001300825458MathSciNetView ArticleGoogle Scholar
 Rockafellar RT: On the virtual convexity of the domain and range of a nonlinear maximal monotone operator. Math. Ann. 1970, 185: 81–90. 10.1007/BF01359698MathSciNetView ArticleGoogle Scholar
 Barbu V: Nonlinear Semigroups and Differential Equations in Banach Spaces. Editura Academiei Republicii Socialiste România, Bucharest; 1976.View ArticleGoogle Scholar
 Takahashi W: Convex Analysis and Approximation of Fixed Points. Yokohama Publishers, Yokohama; 2000. (in Japanese)Google Scholar
 Rockafellar RT: On the maximality of sums of nonlinear monotone operators. Trans. Am. Math. Soc. 1970, 149: 75–88. 10.1090/S00029947197002822725MathSciNetView ArticleGoogle Scholar
 Zălinescu C: Convex Analysis in General Vector Spaces. World Scientific, River Edge; 2002.View ArticleGoogle Scholar
 Rockafellar RT: On the maximal monotonicity of subdifferential mappings. Pac. J. Math. 1970, 33: 209–216. 10.2140/pjm.1970.33.209MathSciNetView ArticleGoogle Scholar
 Brøndsted A, Rockafellar RT: On the subdifferentiability of convex functions. Proc. Am. Math. Soc. 1965, 16: 605–611. 10.1090/S00029939196501781038View ArticleGoogle Scholar
Copyright
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/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.