- Research
- Open Access
- Published:
Approximate solution of zero point problem involving H-accretive maps in Banach spaces and applications
Journal of Inequalities and Applications volume 2019, Article number: 209 (2019)
Abstract
In this manuscript, we introduce two iterative methods for finding the common zeros of two H-accretive mappings in uniformly smooth and uniformly convex Banach spaces. The proposed iterative methods are based on Mann and Halpern iterative methods and viscosity approximation method. Strong convergence results are established for iterative algorithms. Applications based on convex minimization problem, variational inequality problem and equilibrium problem are derived from the main result. Numerical implementation of the main results and application are demonstrated by some examples. Our results extend, generalize, and unify the previously known results given in literature.
1 Introduction
For a real Banach space \(\mathcal{B}\) with a nonempty, convex, and closed subset \(\mathcal{C}\) and for a set-valued mapping \(\varPhi :\mathcal{B}\rightarrow2^{\mathcal{B}}\) whose domain, range, and graph are defined as
respectively, the inclusion problem or zero point problem is to search a point \(\tilde{y}\in\mathcal{B}\) such that
The set of zeros of Φ is defined as
A very effective and convenient approach to solve a large number of problems in optimization theory and nonlinear analysis is their formulation in the zero point problem. In problem (1.1), different mathematicians defined Φ in various ways in various frameworks, such as maximal monotone in Hilbert spaces and m-accretive in Banach spaces, and they introduced iterative schemes to solve (1.1).
In 2004, Fang et al. [12] introduced the notion of H-accretive mappings in Banach spaces. It is noted that H-accretive mapping is a generalization of m-accretive mapping. They defined the resolvent operator affiliated to H-accretive mapping and showed that it is single-valued and Lipschitz continuous.
One of the important methods to solve (1.1) for maximal monotone operator Φ in a Hilbert space \(\mathcal{H}\) is proximal point algorithm which was introduced by Martinet [20] and defined in the following manner:
where \(I:\mathcal{H}\rightarrow\mathcal{H}\) is the identity mapping, \(J_{\mu_{n},\varPhi}(x_{n})=(I+\mu_{n}\varPhi)^{-1}(x_{n})\), and \(\{\mu _{n}\}\subset\mathbb{R}_{+}\).
In 1976, Rockafellar [28] posed an open question regarding the strong convergence of the sequence obtained from (1.2); indeed an example was given by Güler [13] discussing that Rockafellar’s proximal point algorithm does not strongly converge. Furthermore, Bauschke et al. [3] showed by an example that proximal point algorithm only converges weakly but not in norm.
In 2005, a modification of Mann iteration method was presented by Kim et al. [17] for finding the solution of (1.1) with an m-accretive mapping Φ in a uniformly smooth Banach space and described in the following manner:
where \(l\in\mathcal{B}\), \(\{\rho_{n}\}\subset(0,1)\), and \(\{\mu _{n}\}\subset\mathbb{R}_{+}\) with certain conditions. They established an outcome of the strong convergence of the sequence \(\{ x_{n}\}\) acquired from (1.3).
In 2007, the following iteration scheme:
where \(l\in\mathcal{C}\) and the sequences \(\{\kappa_{n}\}\), \(\{\rho _{n}\}\subset(0,1)\), was described by Qin et al. [22] in a uniformly smooth Banach space. The demonstration of the strong convergence of the sequence \(\{x_{n}\}\) acquired from (1.4) to a zero of an m-accretive mapping Φ was also given.
Based on viscosity approximation method, in 2008, Chen et al. [9] developed an iterative scheme in the following manner:
where \(g:\mathcal{C}\rightarrow\mathcal{C}\) is a contraction mapping. They established an outcome which ensures the strong convergence of the sequence \(\{x_{n}\}\) acquired from (1.5) to a zero of an m-accretive mapping Φ in a uniformly smooth Banach space. For detailed reviews regarding iteration methods to various inclusion problems, see [1, 6, 7, 10, 29, 30] and the references therein.
In this manuscript, we consider the problem of searching common zeros of two H-accretive mappings in a uniformly smooth and uniformly convex Banach space, i.e., search \(\tilde{y}\in\mathcal{B}\) such that
where Φ, \(\varPsi:\mathcal{B}\rightarrow2^{\mathcal{B}}\) are H-accretive mappings.
In 2005, Bauschke et al. [2] studied problem (1.6) for two maximal monotone mappings Φ and Ψ in a Hilbert space and developed the following iterative method:
They proved that the sequence \(\{x_{n}\}\) obtained from (1.7) is weakly convergent to a fixed point of \(J_{\mu,\varPhi}J_{\mu,\varPsi}\). This iterative method is based on alternating resolvents method which is an extension of alternating projections method introduced and studied by von Neumann [33] and Bregmann [5]. For other modification of von Neumann’s alternating projections method, see [8].
Further, in 2011, Boikanyo et al. [4] constructed an iterative scheme using alternating resolvents method in the following manner:
where \(l\in\mathcal{H}\), \(\{\rho_{n}\}\subset(0,1)\), and \(\{e_{n}\} \), \(\{e_{n}^{\prime}\}\) are error sequences. Under appropriate restrictions on control sequences, they established the strong convergence of the sequence obtained from (1.8) to a common zero point of maximal monotone mappings Φ and Ψ in a Hilbert space.
Motivated by the above discussed work, we introduce two iterative methods which are based on Mann and Halpern iteration methods and viscosity approximation method. We call it Mann–Halpern type iterative method and Mann type viscosity iterative method. Our proposed iteration methods include alternating resolvents method as a particular case for suitable choice of control sequence. Strong convergence results are established for both iterative methods. Some consequences and applications are derived from the main result. We present numerical examples to show the implementation of our main result and its application. Our results extend, generalize, and unify the results given by Bauschke et al. [2], Kim et al. [17], Qin et al. [22], Chen et al. [9], and Boikanyo et al. [4].
2 Preliminaries
For the whole of this manuscript, let \(\mathcal{B}\) be a Banach space with norm \(\|\cdot\|\) over the field of real numbers \(\mathbb{R}\), and let \(\mathcal{B}^{\ast}\) be its dual space. Let \(\langle\cdot, \cdot\rangle\) be the duality pairing between \(\mathcal{B}\) and \(\mathcal{B}^{\ast}\). We use \(y_{n}\rightarrow y\), \(y_{n}\rightharpoonup y\), and \(y_{n}\overset{\ast}{\rightharpoonup} y\) as symbolic representations for strong, weak, and weak∗ convergence of the sequence \(\{y_{n}\}\) to y in \(\mathcal{B}\) and in \(\mathcal{B}^{\ast}\), respectively. Let \(\mathcal{F}(\mathcal {T})\) be the set of all fixed points of a mapping \(\mathcal{T}\), i.e., \(\mathcal{F}(\mathcal{T})=\{y\in\mathtt{dom}(\mathcal{T}):\mathcal {T}(y)=y\}\). Let \(\mathbb{N}\) and \(\mathbb{R}_{+}\) be the set of all natural numbers and all positive real numbers, respectively.
We take \(\mathcal{J}:\mathcal{B}\rightarrow2^{\mathcal{B}^{\ast}}\) to symbolize the normalized duality mapping, given by
Clearly, if \(\mathcal{B}=\mathcal{H}\), a Hilbert space, then \(\mathcal{J}=I\), where \(I:\mathcal{B}\rightarrow\mathcal{B}\) is the identity mapping. When \(\mathcal{B}\) is smooth, we know from [11] that \(\mathcal{J}\) is single-valued. Now we recollect the following definitions and results which are needed for establishing our main results.
Definition 2.1
Let \(U_{\mathcal{B}}=\{y\in\mathcal{B}:\|y\|=1\}\). The modulus of convexity \(\delta_{\mathcal{B}}:(0,2]\rightarrow[0,1]\) and the modulus of smoothness \(\rho_{\mathcal{B}}:[0,+\infty)\rightarrow [0,+\infty)\) of \(\mathcal{B}\) are defined by
respectively. Then the Banach space \(\mathcal{B}\) is termed
uniformly convex if \(\delta_{\mathcal {B}}(\epsilon)>0\), \(\forall0<\epsilon\leq2\);
uniformly smooth if \(\frac{\rho _{\mathcal{B}}(h)}{h}\rightarrow0\) as \(h\rightarrow0\).
Definition 2.2
A mapping \(\mathcal{T}:\mathcal{B}\rightarrow\mathcal{B}\) is termed
\(\vartheta_{1}\)-Lipschitz continuous if there exists a constant \(\vartheta_{1} \in\mathbb{R}_{+}\) such that
$$\begin{aligned} \bigl\Vert \mathcal{T}(y)-\mathcal{T}(z) \bigr\Vert \leq \vartheta_{1} \Vert y-z \Vert ,\quad \forall y,z\in\mathcal{B}. \end{aligned}$$If \(0<\vartheta_{1}<1\), then \(\mathcal{T}\) is termed \(\vartheta_{1}\)-contraction and if \(\vartheta_{1}=1\), then \(\mathcal{T}\) is termed nonexpansive.
accretive if there exists \(j(y-z)\in \mathcal{J}(y-z)\) such that
$$\begin{aligned} \bigl\langle \mathcal{T}(y)-\mathcal{T}(z),j(y-z)\bigr\rangle \geq0,\quad \forall y,z\in\mathcal{B}; \end{aligned}$$strictly accretive if there exists \(j(y-z)\in\mathcal{J}(y-z)\) such that
$$\begin{aligned} \bigl\langle \mathcal{T}(y)-\mathcal{T}(z),j(y-z)\bigr\rangle \geq0,\quad \forall y,z\in\mathcal{B} \end{aligned}$$and equality holds if and only if \(y=z\);
\(\vartheta_{2}\)-strongly accretive if there exists a constant \(\vartheta_{2}\in\mathbb{R_{+}}\) and \(j(y-z)\in\mathcal{J}(y-z)\) such that
$$\begin{aligned} \bigl\langle \mathcal{T}(y)-\mathcal{T}(z),j(y-z)\bigr\rangle \geq\vartheta _{2} \Vert y-z \Vert ^{2},\quad\forall y,z\in \mathcal{B}. \end{aligned}$$
Definition 2.3
A set-valued mapping \(\varPhi:\mathcal{B}\rightarrow2^{\mathcal{B}}\) is termed
accretive if there exists \(j(y-z)\in \mathcal{J}(y-z)\) such that
$$\begin{aligned} \bigl\langle v-w,j(y-z)\bigr\rangle \geq0,\quad\forall y,z\in\mathcal {B}, v\in \varPhi(y), w\in\varPhi(z); \end{aligned}$$m-accretive if Φ is accretive and \(\mathtt{range}(I+\mu\varPhi)=\mathcal{B}\), \(\forall\mu\in \mathbb{R}_{+}\).
Definition 2.4
A mapping \(J_{\mu,\varPhi}:\mathtt{range}(I+\mu\varPhi)\rightarrow \mathtt{dom}(\varPhi)\) defined by \(J_{\mu,\varPhi}(y)=(I+\mu\varPhi )^{-1}(y)\), \(\mu\in\mathbb{R}_{+}\), is termed resolvent affiliated to an accretive mapping Φ.
Definition 2.5
([12])
Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) be a single-valued mapping. Then a set-valued mapping \(\varPhi:\mathcal{B}\rightarrow 2^{\mathcal{B}}\) is termed H-accretive if the following are satisfied:
Φ is accretive;
\(\mathtt{range}(H+\vartheta_{3}\varPhi)=\mathcal {B}\), where \(\vartheta_{3}\in\mathbb{R}_{+}\).
For further generalizations of H-accretive mapping, see [15, 16].
Remark 2.1
From Definition 2.5, we have the following as special cases:
If \(\mathcal{B}=\mathcal{H}\), a Hilbert space and \(H\equiv I\), then the H-accretive mapping becomes maximal monotone mapping.
If \(H\equiv I\), then the H-accretive mapping becomes m-accretive mapping.
The identity mapping I is m-accretive, but it is not H-accretive mapping, where \(H(y)=y^{2}\) (see [12]).
If we take \(\mathcal{B}=\mathbb {R}\), \(H(y)=-y^{3}\), and
$$\begin{aligned} \mathcal{T}(y)=\operatorname{sgn}(y)= \textstyle\begin{cases} 1,&y>0, \\ 0,&y=0, \\ -1,&y< 0, \end{cases}\displaystyle \end{aligned}$$for all \(y\in\mathcal{B}\), then \(\mathcal{T}\) is H-accretive, but it is not m-accretive (see [19]).
Lemma 2.1
([12])
Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) and \(\varPhi:\mathcal {B}\rightarrow2^{\mathcal{B}}\) be strictly accretive and H-accretive mappings, respectively. Then the operator \((H+\mu\varPhi )^{-1}\), where \(\mu\in\mathbb{R}_{+}\), is single-valued.
Definition 2.6
([12])
Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) and \(\varPhi:\mathcal {B}\rightarrow2^{\mathcal{B}}\) be strictly accretive and H-accretive mappings, respectively. The resolvent operator \(J_{\mu,\varPhi}^{H}:\mathcal{B}\rightarrow\mathcal{B}\) affiliated to H and Φ is defined by
Using Definition 2.6, one can easily prove that
Lemma 2.2
([12])
Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) be a \(\xi_{H}\)-strongly accretive mapping, and let \(\varPhi:\mathcal{B}\rightarrow2^{\mathcal {B}}\) be an H-accretive mapping. Then the resolvent operator \(J_{\mu,\varPhi}^{H}:\mathcal{B}\rightarrow\mathcal{B}\) is \(\frac {1}{\xi_{H}}\)-Lipschitz continuous.
We give an example which legitimizes Lemma 2.2.
Example 2.1
Let \(\mathcal{B}=\mathbb{R}\) with usual norm \(\|y\|=|y|\). Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) and \(\varPhi:\mathcal {B}\rightarrow2^{\mathcal{B}}\) be defined by
respectively. It can be conveniently verified that H is 3-strongly accretive, Φ is accretive, and for \(\mu=2\), \(\mathtt {range}(H+\mu\varPhi)=\mathcal{B}\). Hence, Φ is H-accretive. Therefore the resolvent operator \(J_{\mu,\varPhi}^{H}:\mathcal {B}\rightarrow\mathcal{B}\) affiliated with H and Φ is given by
It is quite easy to verify that the resolvent operator defined in (2.2) is single-valued and \(\frac{1}{3}\)-Lipschitz continuous.
Definition 2.7
Let \(\mathcal{C}\) and \(\mathcal{D}\) be nonempty subsets of a Banach space \(\mathcal{B}\) such that \(\mathcal{C}\) is closed and convex and \(\mathcal{D}\subset\mathcal{C}\). A mapping \(Q:\mathcal {C}\rightarrow\mathcal{D}\) is termed
sunny if \(Q[Q(y)+h(I-Q)y]=Q(y)\), \(\forall y\in\mathcal{C}\) and \(h\geq0\), whenever \(Q(y)+h(I-Q)y\in \mathcal{C}\);
retraction if \(Q(y)=y\), \(\forall y\in \mathcal{D}\);
sunny nonexpansive retraction from \(\mathcal{C}\) onto \(\mathcal{D}\) if Q is a retraction from \(\mathcal{C}\) onto \(\mathcal{D}\), which is also sunny and nonexpansive.
Moreover, \(\mathcal{D}\) is termed sunny nonexpansive retract of \(\mathcal{C}\) if there exists a sunny nonexpansive retraction Q from \(\mathcal{C}\) onto \(\mathcal{D}\) (see [18]).
Proposition 2.3
([23])
Let \(\mathcal{C}\) and \(\mathcal{D}\) be nonempty subsets of a smooth Banach space \(\mathcal{B}\) such that \(\mathcal{C}\) is closed and convex and \(\mathcal{D}\subset\mathcal{C}\). If \(\mathcal{D}\) is a retract of \(\mathcal{C}\) with a retraction \(Q:\mathcal{C}\rightarrow \mathcal{D}\), then the following are equivalent:
- (i)
Q is sunny and nonexpansive;
- (ii)
\(\|Q(y)-Q(z)\|^{2}\leq\langle y-z,\mathcal {J}(Q(y)-Q(z))\rangle\), \(\forall y,z\in\mathcal{C}\);
- (iii)
\(\langle y-Q(y),\mathcal{J}(z-Q(y))\rangle\leq0\), \(\forall y\in\mathcal{C},z\in\mathcal{D}\).
Theorem 2.4
([35])
Let \(\mathcal{C}\) be a nonempty, convex, and closed subset of a uniformly smooth Banach space \(\mathcal{B}\), and let \(W:\mathcal {C}\rightarrow\mathcal{C}\) be a nonexpansive mapping with \(\mathcal {F}(W)\neq\emptyset\). Then the sequence \(\{x_{t}\}\) defined by \(x_{t}=tg(x_{t})+(1-t)W(x_{t})\), where \(g\in\varPi_{\mathcal{C}}\), the set of all contractions on \(\mathcal{C}\) and \(0< t<1\), converges strongly to a point in \(\mathcal{F}(W)\) as \(t\rightarrow0\). If \(Q_{\mathcal{F}(W)}:\varPi_{\mathcal{C}}\rightarrow\mathcal{F}(W)\) is defined as
then \(Q_{\mathcal{F}(W)}(g)\) solves the following variational inequality:
In particular, if \(g=l\in\mathcal{C}\) is a constant, then (2.3) is reduced to the sunny nonexpansive retraction of Reich [25] from \(\mathcal{C}\) onto \(\mathcal{F}(W)\) fulfilling
Proposition 2.5
([19])
Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) be a strongly accretive and Lipschitz continuous mapping with constant \(\xi_{H}\); let \(\varPhi :\mathcal{B}\rightarrow2^{\mathcal{B}}\) be an H-accretive mapping. Then the mappings \(HJ_{\mu,\varPhi}^{H}:\mathcal{B}\rightarrow \mathcal{B}\) and \(J_{\mu,\varPhi}^{H}H:\mathcal{B}\rightarrow\mathcal {B}\) are nonexpansive.
Lemma 2.6
([21])
Let \(\mathcal{B}\) be a Banach space. Then, for all \(y,z\in\mathcal {B}\) and \(j(y+z)\in\mathcal{J}(y+z)\),
Lemma 2.7
([31])
Let \(\{x_{n}\}\) and \(\{w_{n}\}\) be bounded sequences in a Banach space \(\mathcal{B}\), and let \(\{\sigma_{n}\}\subset[0,1]\) with \(0<\liminf_{n\rightarrow\infty}\sigma_{n}\leq\limsup_{n\rightarrow\infty }\sigma_{n}<1\). Assume that \(x_{n+1}=(1-\sigma_{n})w_{n}+\sigma _{n}x_{n}\), \(n\in\mathbb{N}\), and \(\limsup_{n\rightarrow\infty}(\| w_{n+1}-w_{n}\|-\|x_{n+1}-x_{n}\|)\leq0\). Then \(\lim_{n\rightarrow \infty}\|w_{n}-x_{n}\|=0\).
Lemma 2.8
([35])
Assume that \(\{x_{n}\}\) is a sequence in \(\mathbb{R}_{+}\cup\{0\}\) such that
where \(\{\theta_{n}\}\subset(0,1)\) and \(\{\eta_{n}\}\subset\mathbb {R}\) with
- (i)
\(\sum_{n=1}^{\infty}\theta_{n}=\infty\);
- (ii)
\(\limsup_{n\rightarrow\infty}\frac{\eta_{n}}{\theta _{n}}\leq0\) or \(\sum_{n=1}^{\infty}|\eta_{n}|<\infty\).
Then \(\lim_{n\rightarrow\infty}x_{n}=0\).
Lemma 2.9
([34])
Let \(\mathcal{B}\) be a uniformly convex Banach space. Then there exists a function \(f:[0,+\infty) \rightarrow[0,+\infty)\) which is continuous, strictly increasing, and convex with \(f(0)=0\) such that
for all \(y,z\in U_{\mathcal{B}}[0,r]:=\{v\in\mathcal{B}:\|v\|\leq r, r\in\mathbb{R}_{+}\}\) and \(0\leq h\leq1\).
Lemma 2.10
Let \(\mathcal{B}\) be a real uniformly convex and smooth Banach space. Then there exists a function \(f:[0,2r]\rightarrow\mathbb{R}\) which is continuous, strictly increasing, and convex with \(f(0)=0\) such that
3 Main results
In this section, we demonstrate strong convergence of the sequences acquired from the proposed iterative methods for finding a common zero of two H-accretive mappings. First, we prove the following technical lemma.
Lemma 3.1
Let \(H:\mathcal{B}\rightarrow\mathcal{B}\) and \(\varPhi:\mathcal {B}\rightarrow2^{\mathcal{B}}\) be strictly accretive and H-accretive mappings, respectively. Then, for \(\mu,\nu\in\mathbb {R}_{+}\) and \(y\in\mathcal{B}\),
Proof
For \(y\in\mathcal{B}\), let \(\tilde{y}=J_{\mu,\varPhi}^{H}(y)\). This implies that
Then
which implies that
□
Theorem 3.2
Let \(\mathcal{B}\) be a uniformly convex and uniformly smooth Banach space. Let \(H_{1},H_{2}:\mathcal{B}\rightarrow\mathcal{B}\) be single-valued mappings such that \(H_{1}\) is strongly accretive and Lipschitz continuous with constant \(\xi_{H_{1}}\) and \(H_{2}\) is strongly accretive and Lipschitz continuous with constant \(\zeta _{H_{2}}\). Let \(\varPhi,\varPsi:\mathcal{B}\rightarrow2^{\mathcal{B}}\) be \(H_{1}\)-accretive and \(H_{2}\)-accretive mappings, respectively. Assume that \(\varOmega:=\varPhi^{-1}(0)\cap\varPsi^{-1}(0)\neq\emptyset\). Let the sequences \(\{y_{n}\}\) and \(\{x_{n}\}\) be generated by the following iterative scheme:
where \(l\in\mathcal{B}\) is an arbitrary element, sequences \(\{\mu _{n}\}\), \(\{\nu_{n}\}\) are in \(\mathbb{R}_{+}\), and sequences \(\{ \kappa_{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), \(\{\tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau_{n}=1\), \(n\in\mathbb{N}\). Assume that the following conditions are fulfilled:
- (\(\mathrm{C}_{1}\)):
\(\liminf_{n\rightarrow\infty}\kappa _{n}\tau_{n}>0\), \(\lim_{n\rightarrow\infty}|\kappa_{n+1}-\kappa_{n}|=0\);
- (\(\mathrm{C}_{2}\)):
\(\lim_{n\rightarrow\infty}\rho _{n}=0\), \(\sum_{n=1}^{\infty}\rho_{n}=\infty\);
- (\(\mathrm{C}_{3}\)):
\(0<\liminf_{n\rightarrow\infty} \sigma _{n}\leq\limsup_{n\rightarrow\infty} \sigma_{n}<1\);
- (\(\mathrm{C}_{4}\)):
For some \(\epsilon\in\mathbb{R}_{+}\) and for all \(n\in\mathbb{N}\), \(\mu_{n}\geq\epsilon\), \(\nu_{n}\geq \epsilon\) and \(\lim_{n\rightarrow\infty}|\mu_{n+1}-\mu_{n}|=0\) and \(\lim_{n\rightarrow\infty}|\nu_{n+1}-\nu_{n}|=0\).
Then the sequence \(\{x_{n}\}\) converges strongly to \(Q_{\varOmega}l\), where \(Q_{\varOmega}:\mathcal{B}\rightarrow\varOmega\) is a sunny nonexpansive retraction from \(\mathcal{B}\) onto Ω.
Proof
The proof will be divided into six steps.
Step 1. We show that \(\{x_{n}\}\) and \(\{y_{n}\}\) are bounded. Assume that \(w\in\varOmega\). By utilizing (2.1), we find that
Then, from (3.1), (3.2), and Proposition 2.5, we have
Now we calculate \(\|y_{n}-w\|\).
Using (3.4) and the relation \(\rho_{n}+\sigma_{n}+\tau _{n}=1\) in (3.3), we obtain by induction
Hence \(\{x_{n}\}\) is bounded. Therefore \(\{y_{n}\}\), \(\{J_{\mu _{n},\varPhi}^{H_{1}}H_{1}(x_{n})\}\), and \(\{J_{\nu_{n},\varPsi }^{H_{2}}H_{2}(y_{n})\}\) are also bounded.
Step 2. We claim that \(\lim_{n\rightarrow\infty}\| x_{n+1}-x_{n}\|=0\) and \(\lim_{n\rightarrow\infty}\|y_{n+1}-y_{n}\|=0\).
Setting \(x_{n+1}=(1-\sigma_{n})w_{n}+\sigma_{n}x_{n}\), we see that \(w_{n}=\frac{x_{n+1}-\sigma_{n}x_{n}}{1-\sigma_{n}}\). Then
where \(L_{1}=\sup_{n} \{\|l\| + \|J_{\nu_{n+1},\varPsi}^{H_{2}}H_{2}(y_{n+1}) \| \}\). Now, two cases arise.
Case I. When \(\nu_{n+1}\geq\nu_{n}\), by utilizing Lemmas 3.1 and 2.2, we acquire
where \(L_{2}=\frac{1}{\epsilon} [\sup_{n} \{\|y_{n}\| + \|J_{\nu_{n},\varPsi}^{H_{2}}H_{2}(y_{n}) \| \} ]\).
Case II. When \(\nu_{n+1}<\nu_{n}\), by utilizing Lemmas 3.1 and 2.2, we acquire
From (3.5), (3.6), and (3.7), we obtain
By utilizing (3.1) and Lemma 3.1, we acquire
where \(L_{3}=\sup_{n} \{\|x_{n}\|+ \|J_{\mu_{n}, \varPhi }^{H_{1}}H_{1}(x_{n}) \| \}\) and \(L_{4}=\frac {1}{\epsilon} [\sup_{n} \{ \|J_{\mu_{n},\varPhi}^{H_{1}}H_{1}(x_{n}) \|+\|x_{n}\| \} ]\).
Combining (3.8) and (3.9), we acquire
From (\(\mathrm{C}_{1}\)), (\(\mathrm{C}_{2}\)), (\(\mathrm{C}_{4}\)), and (3.10), we acquire
It follows from Lemma 2.7 that \(\lim_{n\rightarrow\infty}\| w_{n}-x_{n}\|=0\), and hence
From (\(\mathrm{C}_{1}\)), (\(\mathrm{C}_{4}\)), (3.9), and (3.11), we acquire
Step 3. Our claim is \(\lim_{n\rightarrow\infty}\| x_{n+1}-y_{n}\|=0\). From (3.1), (3.4), and Lemma 2.10, we have
which implies that
Then, by using (3.1) and (3.12), we acquire
which implies that
By utilizing (3.11), conditions (\(\mathrm{C}_{1}\))–(\(\mathrm {C}_{2}\)), and the property of f in (3.13), we obtain
Since \(\|x_{n+1}-y_{n}\|\leq\|x_{n+1}-x_{n}\|+\|x_{n}-y_{n}\|\), so from (3.11) and (3.14), we acquire that \(\lim_{n\rightarrow\infty}\|x_{n+1}-y_{n}\|=0\).
Step 4. We claim that \(\lim_{n\rightarrow\infty}\| W(x_{n})-x_{n}\|=0\), where \(W=\frac{1}{2} (J_{\vartheta,\varPhi }^{H_{1}}H_{1}+J_{\vartheta,\varPsi}^{H_{2}}H_{2} )\) and \(0<\vartheta<\epsilon\).
First, we compute \(\|x_{n}-J_{\vartheta,\varPhi }^{H_{1}}H_{1}(x_{n}) \|\). From Lemma 3.1 and (3.1), we have
Since \(\|x_{n}-y_{n}\|\rightarrow0\) as \(n\rightarrow\infty\) (from (3.14)), we acquire that \(\lim_{n\rightarrow\infty} \| x_{n}-J_{\vartheta,\varPhi}^{H_{1}}H_{1}(x_{n}) \|=0\).
Next, we compute \(\|x_{n}-J_{\vartheta,\varPsi }^{H_{2}}H_{2}(x_{n}) \|\).
Since \(\|x_{n+1}-x_{n}\|\rightarrow0\) (from (3.11)), \(\| x_{n}-y_{n}\|\rightarrow0\) (from (3.14)), and \(\rho _{n}\rightarrow0\) (from (\(\mathrm{C}_{2}\))) as \(n\rightarrow\infty \), we acquire that \(\lim_{n\rightarrow\infty} \| x_{n}-J_{\vartheta,\varPsi}^{H_{2}}H_{2}(x_{n}) \|=0\). Hence it follows from (3.15) that \(\lim_{n\rightarrow\infty}\| W(x_{n})-x_{n}\|=0\).
Step 5. Our claim is \(\limsup_{n\rightarrow\infty} \langle(I-Q_{\varOmega})l,\mathcal{J}(x_{n}-Q_{\varOmega}l) \rangle \leq0\). Define a sequence \(\{x_{t}\}\) by \(x_{t}=tl+(1-t)W(x_{t})\), \(t\in (0,1)\). Then Theorem 2.4 ensures the strong convergence of \(\{ x_{t}\}\) to \(Q_{\varOmega}l\in\mathcal{F}(W)=\varOmega\). Now
which implies that
Let \(L_{5}=\sup\{\|x_{t}-x_{n}\|:t\in(0,1),n\in\mathbb{N}\}\). Then
Since \(\|W(x_{n})-x_{n}\|\rightarrow0\) as \(n\rightarrow\infty\), we acquire
By utilizing the fact that \(x_{t}\rightarrow Q_{\varOmega}l\) as \(t\rightarrow0^{+}\) and \(\mathcal{J}\) is norm-to-weak∗ uniformly continuous on bounded subsets of \(\mathcal{B}\), we acquire
For \(\varepsilon>0\), there is \(0<\delta<1\) such that
From (3.16), we acquire
Since ε is arbitrary, we acquire that \(\limsup_{n\rightarrow\infty} \langle(I-Q_{\varOmega})l,\mathcal {J}(x_{n}-Q_{\varOmega}l) \rangle \leq0\).
Step 6. Finally, our claim is \(x_{n}\rightarrow Q_{\varOmega}l\) as \(n\rightarrow\infty\). From (3.1), (3.4), Lemmas 2.6 and 2.9, it follows that
Evidently, \(\sum_{n=1}^{\infty}\rho_{n}=\infty\), \(\{\rho_{n}\} \subset(0,1)\) and \(\limsup_{n\rightarrow\infty}\frac{\eta _{n}}{\rho_{n}}\leq0\). Hence, by Lemma 2.8, we acquire that \(x_{n}\rightarrow Q_{\varOmega}l\) as \(n\rightarrow\infty\). Thus, the proof is completed. □
In the next theorem, we prove the strong convergence of the sequence generated by the following Mann type viscosity approximation method:
Theorem 3.3
Let \(\mathcal{B}\) be a uniformly convex and uniformly smooth Banach space. Let \(H_{1},H_{2}:\mathcal{B}\rightarrow\mathcal{B}\) be single-valued mappings such that \(H_{1}\) is strongly accretive and Lipschitz continuous with constant \(\xi_{H_{1}}\) and \(H_{2}\) is strongly accretive and Lipschitz continuous with constant \(\zeta _{H_{2}}\). Let \(\varPhi, \varPsi:\mathcal{B}\rightarrow2^{\mathcal{B}}\) be \(H_{1}\)-accretive and \(H_{2}\)-accretive mappings, respectively; let \(g:\mathcal{B}\rightarrow\mathcal{B}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=\varPhi ^{-1}(0)\cap\varPsi^{-1}(0)\neq\emptyset\). Let the sequences \(\{v_{n}\} \) and \(\{u_{n}\}\) be generated by (3.17). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm {C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{ u_{n}\}\) converges strongly to \(\tilde{u}=Q_{\varOmega}g(\tilde{u})\in \varOmega\).
Proof
Assume that ũ is a unique fixed point of \(Q_{\varOmega}g\). Then \(Q_{\varOmega}g(\tilde{u})=\tilde{u}\). If we put \(g(\tilde{u})\) in place of l in (3.1), then Theorem 3.2 ensures the strong convergence of \(\{x_{n}\}\) to \(Q_{\varOmega}g(\tilde{u})=\tilde {u}\), i.e., \(\lim_{n\rightarrow\infty}\|x_{n}-\tilde{u}\|=0\).
First, we demonstrate that \(\lim_{n\rightarrow\infty}\|u_{n}-x_{n}\| =0\). We are assuming on the contrary that
Then we can pick ε such that \(0<\varepsilon<\limsup_{n\rightarrow\infty}\|u_{n}-x_{n}\|\). As \(\{x_{n}\}\) is strongly convergent to ũ, so there exists \(m^{\prime}\in\mathbb {N}\) such that
Now two possibilities arise.
- (P1)
There exists \(m\in\mathbb{N}\) with \(m\geq m^{\prime}\) and \(\|u_{m}-x_{m}\|\leq\varepsilon\);
- (P2)
\(\|u_{n}-x_{n}\|>\varepsilon\), \(\forall n\geq m^{\prime}\).
In possibility (P1),
By induction,
which is a contradiction to \(\varepsilon< \limsup_{n\rightarrow \infty}\|u_{n}-x_{n}\|\).
In possibility (P2), for all \(n\geq m^{\prime}\), we acquire
By Lemma 2.8, we acquire that \(\lim_{n\rightarrow\infty}\| u_{n}-x_{n}\|=0\), which is a contradiction. Therefore \(\lim_{n\rightarrow\infty}\|u_{n}-x_{n}\|=0\), and hence
Thus, the proof is completed. □
4 Consequences
In this section, we deduce some consequences from our main results.
Corollary 4.1
Let \(\mathcal{B}\) be a uniformly convex and uniformly smooth Banach space. Let \(\varPhi,\varPsi:\mathcal{B}\rightarrow2^{\mathcal{B}}\) be m-accretive mappings, and let \(g:\mathcal{B}\rightarrow\mathcal {B}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=\varPhi^{-1}(0)\cap\varPsi^{-1}(0)\neq\emptyset\). Let the sequences \(\{v_{n}\}\) and \(\{u_{n}\}\) be generated by the following iterative scheme:
where \(J_{\mu_{n},\varPhi}(u_{n})=(I+\mu_{n}\varPhi)^{-1}(u_{n})\), \(J_{\nu_{n},\varPsi}(v_{n})=(I+\nu_{n}\varPsi)^{-1}(v_{n})\), sequences \(\{ \mu_{n}\}\), \(\{\nu_{n}\}\) are in \(\mathbb{R}_{+}\) and \(\{\kappa _{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), \(\{\tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau_{n}=1\), \(n\in\mathbb{N}\). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde {u}=Q_{\varOmega}g(\tilde{u})\in\varOmega\).
Proof
If we take \(H_{1}\equiv H_{2}\equiv I\) in Theorem 3.3, then the conclusion holds. □
In the framework of Hilbert spaces, we have the following outcomes.
Corollary 4.2
Let \(\mathcal{H}\) be a Hilbert space. Let \(H_{1},H_{2}:\mathcal {B}\rightarrow\mathcal{B}\) be single-valued mappings such that \(H_{1}\) is strongly monotone and Lipschitz continuous with constant \(\xi_{H_{1}}\) and \(H_{2}\) is strongly monotone and Lipschitz continuous with constant \(\zeta_{H_{2}}\). Let \(\varPhi,\varPsi:\mathcal {H}\rightarrow2^{\mathcal{H}}\) be \(H_{1}\)-monotone and \(H_{2}\)-monotone mappings, respectively, and let \(g:\mathcal {H}\rightarrow\mathcal{H}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=\varPhi^{-1}(0)\cap\varPsi ^{-1}(0)\neq\emptyset\). Let the sequences \(\{v_{n}\}\) and \(\{u_{n}\}\) be generated by the iterative scheme (3.17). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde {u}=P_{\varOmega}g(\tilde{u})\in\varOmega\), where \(P_{\varOmega}:\mathcal {H}\rightarrow\varOmega\) is a metric projection from \(\mathcal{H}\) onto Ω.
For \(H\equiv I\), the H-monotone mapping becomes maximal monotone mapping in Hilbert spaces. Hence we have the following outcome.
Corollary 4.3
Let \(\mathcal{H}\) be a Hilbert space. Let \(\varPhi,\varPsi:\mathcal {H}\rightarrow2^{\mathcal{H}}\) be maximal monotone mappings, and let \(g:\mathcal{H}\rightarrow\mathcal{H}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=\varPhi ^{-1}(0)\cap\varPsi^{-1}(0)\neq\emptyset\). Let the sequences \(\{v_{n}\} \) and \(\{u_{n}\}\) be generated by the iterative scheme (4.1). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde{u}=P_{\varOmega}g(\tilde{u})\in\varOmega\).
5 Applications
In this section, firstly, we examine a few of applications based on convex minimization problem.
Theorem 5.1
Let \(\mathcal{H}\) be a Hilbert space. Let \(f_{1},f_{2}:\mathcal {H}\rightarrow(-\infty,+\infty]\) be proper mappings with convexity and lower semi-continuity, and let \(g:\mathcal{H}\rightarrow\mathcal {H}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=(\partial f_{1})^{-1}(0)\cap(\partial f_{2})^{-1}(0)\neq\emptyset\). Let the sequences \(\{v_{n}\}\), \(\{ w_{n}\}\), and \(\{u_{n}\}\) be generated by the following iterative scheme:
where sequences \(\{\mu_{n}\}\), \(\{\nu_{n}\}\) are in \(\mathbb{R}_{+}\) and sequences \(\{\kappa_{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), \(\{ \tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau _{n}=1\), \(n\in\mathbb{N}\). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde {u}=P_{\varOmega}g(\tilde{u})\in\varOmega\).
Proof
From (5.1), we can write
For a proper mapping with convexity and lower semi-continuity, the subdifferential mapping is maximal monotone (see [26]). Thus \(\partial f_{1}\) and \(\partial f_{2}\) are maximal monotone in \(\mathcal {H}\). Hence (5.2) is equivalent to
that is,
and
is equivalent to
Therefore,
With the assistance of Corollary 4.3, including \(\varPhi=\partial f_{1}\) and \(\varPsi=\partial f_{2}\), we get the strong convergence of \(\{ u_{n}\}\) to a point \(\tilde{u}\in\varOmega\), which is the conclusion. □
Corollary 5.2
Let \(\mathcal{H}\) be a Hilbert space. Let \(f:\mathcal{H}\rightarrow (-\infty,+\infty]\) be a proper mapping with convexity and lower semi-continuity, and let \(g:\mathcal{H}\rightarrow\mathcal{H}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=(\partial f)^{-1}(0)\neq\emptyset\). Let the sequences \(\{ v_{n}\}\) and \(\{u_{n}\}\) be generated by the following iterative scheme:
where \(\{\mu_{n}\}\subset\mathbb{R}_{+}\) and sequences \(\{\kappa _{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), \(\{\tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau_{n}=1\), \(n\in\mathbb {N}\). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde{u}=P_{\varOmega}g(\tilde{u})\in\varOmega\).
For a nonempty, convex, and closed subset \(\mathcal{C}\) of a Hilbert space \(\mathcal{H}\), the indicator function symbolized by \(\mathcal {I}_{\mathcal{C}}\) is defined by
and the normal cone for \(\mathcal{C}\) symbolized by \(\mathcal {N}_{\mathcal{C}}\) at \(y\in\mathcal{C}\) is defined by
Theorem 5.3
Let \(\mathcal{C}_{1}\) and \(\mathcal{C}_{2}\) be nonempty, convex, and closed subsets of a Hilbert space \(\mathcal{H}\). Let \(g:\mathcal {H}\rightarrow\mathcal{H}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Let \(P_{\mathcal{C}_{1}}:\mathcal {H}\rightarrow\mathcal{C}_{1}\) and \(P_{\mathcal{C}_{2}}:\mathcal {H}\rightarrow\mathcal{C}_{2}\) be projections. Assume that \(\varOmega :=\mathcal{C}_{1}\cap\mathcal{C}_{2}\neq\emptyset\). Let the sequences \(\{v_{n}\}\) and \(\{u_{n}\}\) be generated by the following iterative scheme:
where sequences \(\{\kappa_{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), and \(\{\tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau _{n}=1\), \(n\in\mathbb{N}\). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{3}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde {u}=P_{\varOmega}g(\tilde{u})\in\varOmega\).
Proof
Since indicator functions \(\mathcal{I}_{\mathcal{C}_{1}}\) and \(\mathcal{I}_{\mathcal{C}_{2}}\) of \(\mathcal{C}_{1}\) and \(\mathcal {C}_{2}\), respectively, are proper mappings with convexity and lower semi-continuity and \(\operatorname*{argmin}\mathcal{I}_{\mathcal{C}_{1}}=\mathcal {C}_{1}\) and \(\operatorname*{argmin}\mathcal{I}_{\mathcal{C}_{2}}=\mathcal {C}_{2}\), therefore
With the assistance of Theorem 5.1 including \(f_{1}=\mathcal {I}_{\mathcal{C}_{1}}\) and \(f_{2}=\mathcal{I}_{\mathcal{C}_{2}}\) and for all convex closed subset \(\mathcal{C}\) in \(\mathcal{H}\) and for all \(\mu\in\mathbb{R}_{+}\), \(P_{\mathcal{C}}=(I+\mu\partial \mathcal{I}_{\mathcal{C}})^{-1}=(I+\mu\mathcal{N}_{\mathcal {C}})^{-1}\), we secure the conclusion. □
Next, we evaluate a common solution of the variational inequality problem and convex minimization problem.
For a nonempty, convex, and closed subset \(\mathcal{C}\) of a Hilbert space \(\mathcal{H}\) and for a single-valued monotone and hemi-continuous mapping \(\varPhi: \mathcal{C}\rightarrow\mathcal{H}\), the variational inequality problem is to seek a point \(y\in\mathcal {C}\) such that the following inequality is fulfilled:
\(\mathtt{VI}(\mathcal{C},\varPhi)\) stands for the solution set of the variational inequality problem (5.5).
Theorem 5.4
Let \(\mathcal{C}\) be a nonempty, convex, and closed subset of a Hilbert space \(\mathcal{H}\). Let \(f:\mathcal{H}\rightarrow(-\infty ,+\infty]\) be a proper mapping with convexity and lower semi-continuity, and let \(\varPhi:\mathcal{C}\rightarrow\mathcal{H}\) be a monotone and hemi-continuous mapping. Assume that \(\varOmega :=\partial f^{-1}(0)\cap\mathtt{VI}(\mathcal{C},\varPhi)\neq\emptyset \). Let the sequences \(\{v_{n}\}\), \(\{w_{n}\}\), and \(\{u_{n}\}\) be generated by the following iterative scheme:
where sequences \(\{\mu_{n}\}\), \(\{\nu_{n}\}\) are in \(\mathbb{R}_{+}\) and sequences \(\{\kappa_{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), \(\{ \tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau _{n}=1\), \(n\in\mathbb{N}\). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde {u}=P_{\varOmega}g(\tilde{u})\in\varOmega\).
Proof
Define a mapping \(\mathcal{T}\subset\mathcal{H}\times\mathcal{H}\) by
From [27], we know that \(\mathcal{T}\) is maximal monotone, and \(\mathcal{T}^{-1}(0)=\mathtt{VI}(\mathcal{C},\varPhi)\).
From (5.5) and (5.6), we can easily observe that
which gives
With the assistance of Corollary 4.3 and the proof of Theorem 5.1, we obtain the conclusion. □
Finally, we derive an application to equilibrium problem which is as follows:
Let \(\mathcal{C}\) be a nonempty, convex, and closed subset of a Hilbert space \(\mathcal{H}\). Let \(\varTheta:\mathcal{C}\times\mathcal {C}\rightarrow\mathbb{R}\) be a bi-function. The equilibrium problem is to seek a point \(y\in\mathcal{C}\) such that the following inequality is fulfilled:
Let the solution set of equilibrium problem be identified by \(\mathtt {EP}(\varTheta)\). To study problem (5.7), suppose that the following conditions are fulfilled by Θ:
- (\(\mathrm{E}_{1}\)):
For all \(z\in\mathcal{C}\), \(\varTheta(z,z)=0\);
- (\(\mathrm{E}_{2}\)):
Θ is monotone, i.e., for all \(y,z\in\mathcal{C}\), \(\varTheta(y,z)+\varTheta(z,y)\leq0\);
- (\(\mathrm{E}_{3}\)):
\(\lim_{t\downarrow0}\varTheta ((1-t)y+tw,z)\leq\varTheta(y,z)\) for each \(y,z,w\in\mathcal{C}\);
- (\(\mathrm{E}_{4}\)):
For each \(y\in\mathcal{C}\), \(z\longmapsto\varTheta(y,z)\) is convex and lower semi-continuous.
Lemma 5.5
([32])
Let \(\mathcal{C}\) be a nonempty, convex, and closed subset of a Hilbert space \(\mathcal{H}\). Let \(\varTheta:\mathcal{C}\times\mathcal {C}\rightarrow\mathbb{R}\) be a bi-function which fulfills (\(\mathrm{E}_{1}\))–(\(\mathrm{E}_{4}\)). Then a set-valued mapping \(A_{\varTheta}:\mathcal{H}\rightarrow2^{\mathcal{H}}\) defined by
is maximal monotone with \(\mathtt{dom}(A_{\varTheta})\subset\mathcal {C}\), \(\mathtt{EP}(\varTheta)=A_{\varTheta}^{-1}(0)\), and the resolvent \(\mathcal{T}_{\mu,A_{\varTheta}}=(I+\mu A_{\varTheta})^{-1}\) affiliated with \(A_{\varTheta}\) is defined by
Theorem 5.6
Let \(\mathcal{C}\) be a nonempty, convex, and closed subset of a Hilbert space \(\mathcal{H}\). Let \(\varTheta_{1},\varTheta_{2}:\mathcal {C}\times\mathcal{C}\rightarrow\mathbb{R}\) be bi-functions which fulfill (\(\mathrm{E}_{1}\))–(\(\mathrm{E}_{4}\)), and let \(g:\mathcal {H}\rightarrow\mathcal{H}\) be a ϑ-contraction mapping, where \(0< \vartheta<1\). Assume that \(\varOmega:=\mathtt{EP}(\varTheta _{1})\cap\mathtt{EP}(\varTheta_{2})\neq\emptyset\). Let the sequences \(\{v_{n}\}\) and \(\{u_{n}\}\) be generated by the following iterative scheme:
where sequences \(\{\mu_{n}\}\), \(\{\nu_{n}\}\) are in \(\mathbb{R}_{+}\) and sequences \(\{\kappa_{n}\}\), \(\{\rho_{n}\}\), \(\{\sigma_{n}\}\), \(\{ \tau_{n}\}\) are in \([0,1]\) with \(\rho_{n}+\sigma_{n}+\tau_{n}=1\), \(n\in\mathbb {N}\). Assume that conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Then the sequence \(\{u_{n}\}\) converges strongly to \(\tilde{u}=P_{\varOmega}g(\tilde{u})\in\varOmega\).
Proof
With the assistance of Corollary 4.3 and Lemma 5.5, we obtain the conclusion. □
6 Numerical implementation
Firstly, we discuss a numerical example to demonstrate the implementation of iterative schemes (3.1) and (3.17) in Theorem 3.2 and Theorem 3.3, respectively. Matlab R2012a is used for writing all the codes, and it is running on Lenovo Core(TM) i5-2320 CPU @ 3.00 GHz with 4 GB RAM.
Example 6.1
Let \(\mathcal{B}=\mathbb{R}\), with the usual norm \(|\cdot|\). Let us define \(\varPhi,\varPsi:\mathbb{R}\rightarrow2^{\mathbb{R}}\) by \(\varPhi (y)=\{3y\}\) and \(\varPsi(y)=\{4y\}\), and \(H_{1},H_{2}:\mathbb {R}\rightarrow\mathbb{R}\) by \(H_{1}(y)=2y\) and \(H_{2}(y)=\frac {y}{3}\) for all \(y\in\mathbb{R}\). Easily, we can check that \(H_{1}\) is strongly accretive and Lipschitz continuous with constant 2; \(H_{2}\) is strongly accretive and Lipschitz continuous with constant \(\frac {1}{3}\); Φ is \(H_{1}\)-accretive and Ψ is \(H_{2}\)-accretive. We observe that \(\varOmega=\varPhi^{-1}(0)\cap\varPsi^{-1}(0)=\{0\}\). On setting, for all \(n\in\mathbb{N}\), \(\kappa_{n}=\frac{3n+2}{6n}\), \(\mu _{n}=\frac{2n+3}{n+1}=\nu_{n}\), \(\rho_{n}=\frac{1}{n+1}\), \(\sigma_{n}=\frac{1}{3}\), \(\tau_{n}=\frac{2n-1}{3(n+1)}\), all conditions (\(\mathrm{C}_{1}\))–(\(\mathrm{C}_{4}\)) of Theorem 3.2 are fulfilled. Now we consider two cases:
Case I. For the iterative scheme (3.1) of Theorem 3.2, let \(l=0.01\). Then the sequence \(\{x_{n}\}\) obtained from (3.1) strongly converges to \(Q_{\varOmega}(0.01)=0\). For \(x_{1}=-0.4\), the convergence of \(\{x_{n}\}\) and behavior of error \(\|x_{n+1}-x_{n}\|\) are numerically shown in Table 1 and graphically in Figures 1a and 1b, respectively.
Example 6.1, Case I with \(x_{1}=-0.4\)
Case II. For the iterative scheme (3.17) of Theorem 3.3, let \(g(y)=\frac{y}{4}\) for all \(y\in\mathbb {R}\). Then Table 1 and Fig. 2a ensures the strong convergence of the sequence \(\{ u_{n}\}\) obtained from (3.17) to \(\tilde{u}=Q_{\varOmega }(g(\tilde{u}))=0\) for \(u_{1}=0.5\). The graphical representation of error \(\|u_{n+1}-u_{n}\|\) is displayed in Fig. 2b.
Example 6.1, Case II with \(u_{1}=0.5\)
Next, we discuss another numerical example for the iterative scheme (5.1) in Theorem 5.1.
Example 6.2
Let \(\mathcal{H}=\mathbb{R}^{3}\), with the inner product defined by \(\langle y,z \rangle=y_{1}z_{1}+y_{2}z_{2}+y_{3}z_{3}\), \(\forall y=(y^{1},y^{2},y^{3}),z=(z^{1},z^{2},z^{3})\in\mathbb{R}^{3}\) and induced Euclidean norm. For \(k=1,2\), define \(f_{k}:\mathbb {R}^{3}\rightarrow(-\infty,+\infty]\) as follows:
with
It can be easily demonstrated that \(f_{1}\) and \(f_{2}\) are proper mappings with convexity and continuity. In Theorem 5.1, we seek a point \(\tilde{u}\in\varOmega=(\partial f_{1})^{-1}(0)\cap(\partial f_{2})^{-1}(0)\), equivalently, \(\tilde{u}\in\varOmega=\operatorname*{argmin}_{y\in \mathbb{R}^{3}}f_{1}(y)\cap \operatorname*{argmin}_{y\in\mathbb{R}^{3}}f_{2}(y)\), where \(f_{1},f_{2}:\mathbb{R}^{3}\rightarrow(-\infty,+\infty]\) are proper mappings with convexity and lower semi-continuity. For \(f_{k}\), \(k=1,2\), as defined in (6.1),
By taking, for all \(n\in\mathbb{N}\), \(\kappa_{n}=\frac {9n+8}{72n}\), \(\mu_{n}=\frac{2n+3}{n+1}=\nu_{n}\), \(\rho_{n}=\frac{1}{n+1}\), \(\sigma_{n}=\frac{1}{9}\), \(\tau_{n}=\frac{8n-1}{9(n+1)}\), it can be easily verified that all conditions (\(\mathrm{C}_{1}\))–(\(\mathrm {C}_{4}\)) of Theorem 5.1 are fulfilled. Now two cases may occur:
Case I. If \(g(y)=(-0.5,1,0.5)\) for all \(y\in\mathbb {R}^{3}\), then the sequence \(\{u_{n}\}\) obtained from (5.1) strongly converges to \(\tilde{u}= P_{\varOmega }(-0.5,1,0.5)=(0,0.75,0.25)\) for \(u_{1}=(-2,1,2)\). The numerical and graphical demonstration of convergence of the sequence \(\{u_{n}\}\) and the behavior of error \(\|u_{n+1}-u_{n}\|_{2}\) are displayed in Table 2 and Fig. 3, respectively.
Example 6.2, Case I with \(u_{1}=(-2,1,2)\)
Case II. If \(g(y)=\frac{y+1}{16}\) for all \(y\in \mathbb{R}^{3}\), then the sequence \(\{u_{n}\}\) obtained from (5.1) strongly converges to \(\tilde{u}=P_{\varOmega}g(\tilde {u})=(0,0.5,0.5)\) for \(u_{1}=(0,1,2)\). The convergence of \(\{u_{n}\}\) and the behavior of error \(\|u_{n+1}-u_{n}\|_{2}\) are presented numerically in Table 2 and graphically in Fig. 4.
Example 6.2, Case II with \(u_{1}=(0,1,2)\)
7 Conclusion
In this manuscript, we have introduced two iterative methods for finding the common zeros of two H-accretive mappings in the framework of Banach spaces. These iterative methods are based on Mann and Halpern iterative methods and viscosity approximation method. It is easy to observe that for \(\kappa_{n}=0\), \(n\in\mathbb{N}\), our proposed iterative methods consist of method of alternating resolvents, and hence our work extends the methods developed by Bauschke et al. [2], Boikanyo et al. [4], and Liu et al. [19]. We have demonstrated the strong convergence results of the sequences generated by our proposed iterations. Further, we have examined some applications which are based on convex minimization problem, variational inequality problem, and equilibrium problem. Finally, we have presented some numerical examples for implementation of our main results and application.
References
Agarwal, R.P., O’Regan, D., Sahu, D.R.: Fixed Point Theory for Lipschitzian-Type Mappings with Applications. Springer, New York (2009)
Bauschke, H.H., Combettes, P.L., Reich, S.: The asymptotic behavior of the composition of two resolvents. Nonlinear Anal. 60(2), 283–301 (2005)
Bauschke, H.H., Matoušková, E., Reich, S.: Projection and proximal point methods: convergence results and counterexamples. Nonlinear Anal. 56(5), 715–738 (2004)
Boikanyo, O.A., Moroşanu, G.: On the method of alternating resolvents. Nonlinear Anal. 74(15), 5147–5160 (2011)
Bregmann, L.M.: The method of successive projection for finding a common point of convex sets. Sov. Math. Dokl. 162(3), 688–692 (1965)
Ceng, L.C., Petruşel, A., Wong, M.M.: Hybrid viscosity iterative approximation of zeros of m-accretive operators in Banach spaces. Taiwan. J. Math. 15(6), 2459–2481 (2011)
Ceng, L.C., Xu, H.K.: Strong convergence of a hybrid viscosity approximation method with perturbed mappings for nonexpansive and accretive operators. Taiwan. J. Math. 11(3), 661–682 (2007)
Censor, Y., Gibali, A., Reich, S.: A von Neumann alternating method for finding common solutions to variational inequalities. Nonlinear Anal. 75(12), 4596–4603 (2012)
Chen, R., Zhu, Z.: Viscosity approximation method for accretive operator in Banach space. Nonlinear Anal. 69(4), 1356–1363 (2008)
Cholamjiak, P.: A generalized forward-backward splitting method for solving quasi inclusion problems in Banach spaces. Numer. Algorithms 71(4), 915–932 (2016)
Cioranescu, I.: Geometry of Banach Spaces, Duality Mappings and Nonlinear Problems. Kluwer Academic, Dordrecht (1990)
Fang, Y.P., Huang, N.J.: H-accretive operators and resolvent operator technique for solving variational inclusions in Banach spaces. Appl. Math. Lett. 17(6), 647–653 (2004)
Güler, O.: On the convergence of the proximal point algorithm for convex minimization. SIAM J. Control Optim. 29(2), 403–419 (1991)
Kamimura, S., Takahashi, W.: Strong convergence of a proximal-type algorithm in a Banach space. SIAM J. Optim. 13(3), 938–945 (2002)
Kazmi, K.R., Khan, F.A.: Iterative approximation of a solution of multi-valued variational-like inclusion in Banach spaces: A P-η-proximal-point mapping approach. J. Math. Anal. Appl. 325(1), 665–674 (2007)
Kazmi, K.R., Khan, F.A.: Iterative approximation of a unique solution of a system of variational-like inclusions in real q-uniformly smooth Banach spaces. Nonlinear Anal. 67(3), 917–929 (2007)
Kim, T.H., Xu, H.K.: Strong convergence of modified Mann iterations. Nonlinear Anal. 61(1–2), 51–60 (2005)
Kopecká, E., Reich, S.: Nonexpansive retracts in Banach spaces. Banach Cent. Publ. 77, 161–174 (2007)
Liu, S., He, H.: Approximating solution of \(0\in T(x)\) for an H-accretive operator in Banach spaces. J. Math. Anal. Appl. 385(1), 466–476 (2012)
Martinet, B.: Régularisation d’inéquations variationnelles par approximations successives. Rev. Fr. Autom. Inform. Rech. Opér. 4, 154–158 (1970)
Petryshyn, W.V.: A characterization of strict convexity of Banach spaces and other uses of duality mappings. J. Funct. Anal. 6, 282–291 (1970)
Qin, X., Su, Y.: Approximation of a zero point of accretive operator in Banach spaces. J. Math. Anal. Appl. 329(1), 415–424 (2007)
Reich, S.: Asymptotic behavior of contractions in Banach spaces. J. Math. Anal. Appl. 44, 57–70 (1973)
Reich, S.: Constructive techniques for accretive and monotone operators. In: Applied Nonlinear Analysis, Proc. Third Internat. Conf., Univ. Texas, Arlington, Tex., 1978, pp. 335–345. Academic Press, New York (1979)
Reich, S.: Strong convergence theorems for resolvents of accretive operators in Banach spaces. J. Math. Anal. Appl. 75(1), 287–292 (1980)
Rockafellar, R.T.: Characterization of the subdifferentials of convex functions. Pac. J. Math. 17, 497–510 (1966)
Rockafellar, R.T.: On the maximality of sums of nonlinear monotone operators. Trans. Am. Math. Soc. 149, 75–88 (1970)
Rockafellar, R.T.: Monotone operators and the proximal point algorithm. SIAM J. Control Optim. 14(5), 877–898 (1976)
Salahuddin, Ahmad, M.K., Siddiqi, A.H.: Existence results for generalized nonlinear variational inclusions. Appl. Math. Lett. 18(8), 859–864 (2005)
Sunthrayuth, P., Cholamjiak, P.: Iterative methods for solving quasi-variational inclusion and fixed point problem in q-uniformly smooth Banach spaces. Numer. Algorithms 78(4), 1019–1044 (2018)
Suzuki, T.: Strong convergence of Krasnoselskii and Mann’s type sequences for one-parameter nonexpansive semigroups without Bochner integrals. J. Math. Anal. Appl. 305(1), 227–239 (2005)
Takahashi, S., Takahashi, W., Toyoda, M.: Strong convergence theorems for maximal monotone operators with nonlinear mappings in Hilbert spaces. J. Optim. Theory Appl. 147(1), 27–41 (2010)
von Neumann, J.: Functional Operators II: The Geometry of Orthogonal Spaces. Princeton University Press, Princeton (1950)
Xu, H.K.: Inequalities in Banach spaces with applications. Nonlinear Anal. 16(12), 1127–1138 (1991)
Xu, H.K.: Viscosity approximation methods for nonexpansive mappings. J. Math. Anal. Appl. 298(1), 279–291 (2004)
Acknowledgements
The authors are thankful to the editor and anonymous referees for their helpful comments and suggestions.
Availability of data and materials
Not applicable.
Funding
No funding available.
Author information
Authors and Affiliations
Contributions
The authors contributed equally and significantly in writing this manuscript. All the authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Vaish, R., Sarfaraz, M., Ahmad, M.K. et al. Approximate solution of zero point problem involving H-accretive maps in Banach spaces and applications. J Inequal Appl 2019, 209 (2019). https://doi.org/10.1186/s13660-019-2163-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13660-019-2163-y
MSC
- 47H04
- 47H09
- 47H10
- 47J25
Keywords
- H-accretive mappings
- Alternating resolvents
- Mann’s iteration
- Halpern iteration
- Viscosity approximation method
- Banach space