Abstract
In this paper, we introduce a new inertial Tseng’s extragradient method with self-adaptive step sizes for approximating a common solution of split equalities of equilibrium problem (EP), non-Lipschitz pseudomonotone variational inequality problem (VIP) and fixed point problem (FPP) of nonexpansive semigroups in real Hilbert spaces. We prove that the sequence generated by our proposed method converges strongly to a common solution of the EP, pseudomonotone VIP and FPP of nonexpansive semigroups without any linesearch procedure nor the sequential weak continuity condition often assumed by authors when solving non-Lipschitz VIPs. Finally, we provide some numerical experiments for the proposed method in comparison with related methods in the literature. Our result improves, extends and generalizes several of the existing results in this direction.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(\mathcal {C}\) be a nonempty, closed and convex subset of a real Hilbert space \(\mathcal {H}.\) The variational inequality problem (VIP) is defined as follows: Find \(x\in \mathcal {C}\) such that
where \(A:\mathcal {H}\rightarrow \mathcal {H}\) is an operator. We denote by \(VI(\mathcal {C},A)\) the solution set of the problem (1.1).
Definition 1.1
Let \(A:\mathcal {H}\rightarrow \mathcal {H}\) be a mapping. Then, A is said to be
-
(i)
L-co-coercive (or L-inverse strongly monotone), if there exists a constant \(L>0\) such that
$$\begin{aligned} \big <Ax-Ay,x-y\big >\ge L\Vert Ax-Ay\Vert ^2,~~ \forall x,y \in \mathcal {H}, \end{aligned}$$ -
(ii)
Monotone, if
$$\begin{aligned}\big <Ax-Ay,x-y\big >\ge 0,~~ \forall x,y \in \mathcal { H}. \end{aligned}$$ -
(iii)
Pseudomonotone, if
$$\langle Ay, x-y \rangle \ge 0 \implies ~\langle Ax,x-y \rangle \ge 0,~\forall x,y \in \mathcal {H},$$
Note that (i) \(\implies \) (ii) \(\implies \) (iii) but the converses are not always true.
A central problem in nonlinear analysis is the VIP, which was first introduced independently by Fichera [18] and Stampacchia [51]. It plays an important role in the study of several important concepts in pure and applied sciences such as mechanics, neccessary optimality conditions, operations research, systems of nonlinear equations, among others (see [19, 25, 62]). Many authors have analyzed and studied iterative algorithms for approximating the solution of the VIP (1.1) and other related optimization problems, (see [2, 10, 20, 27, 36, 41, 52, 56], and the references therein).
Under certain conditions, there are two common methods used in approximating the solution of the VIP (1.1). These methods are the projection method and the regularized method. To use these methods, a certain level of monotonicity is required for the cost operator. In this work, our main focus is on the projection method. Several authors have proposed and studied projection type algorithms for approximating the solutions of VIP (1.1) (see [1, 13, 14, 22, 30, 43, 60] and other references therein).
Tseng [57] introduced and studied Tseng’s extragradient method for approximating the solution of the VIP (1.1). The proposed method is defined as follows:
where A is monotone, L-Lipschitz continuous and \(\lambda \in \Big (0,\frac{2}{L}\Big ).\) The author obtained a weak convergence result under the assumption that \(VI(\mathcal {C},A)\ne \emptyset .\)
The equilibrium problem (EP) was introduced by Blum and Oettli [7] and they defined it as follows: Find \(x\in \mathcal {C}\) such that
where \(\mathcal {C}\) is a nonempty, closed and convex subset of a real Hilbert space \(\mathcal {H}\) and \(\Phi :\mathcal {C}\times \mathcal {C}\rightarrow \mathbb {R}\) is a bifunction. A point \(x\in \mathcal {C}\) that solves this problem is called the equilibrium point. We denote the solution set of EP (1.2) by \(EP(\Phi ).\) The EP (1.2) has received a lot of attention from several authors due to its application to problems arising in the field of optimization, economics, physics, variational inequalities, among others (see, for example, [39, 42, 47, 53] and other references therein). Several authors have analyzed and proposed various iterative algorithms for approximating the solution of the EP and other related optimization problems, (see, for example, [24, 40, 46] and other references therein).
Let \(\mathcal {H}_1, \mathcal {H}_2\) and \(\mathcal {H}_3\) be real Hilbert spaces. Let \(\mathcal {C},\mathcal {Q}\) be nonempty, closed and convex subsets of \(\mathcal {H}_1\) and \(\mathcal {H}_2,\) respectively. Let \(\mathcal {F}_1:\mathcal {H}_1\rightarrow \mathcal {H}_3\) and \(\mathcal {F}_2:\mathcal {H}_2\rightarrow \mathcal {H}_3\) be bounded linear operators. The split equality problem (SEP) is defined as follows:
The SEP which was first proposed by Moudafi [37] allows asymmetric and partial relations between the variables x and y. It is used in numerous practical problems such as game theory, medical image reconstruction, partial differential equation, decomposition method, among others (see [29, 50]). We denote the solution set of (1.3) by
Several authors have studied several effective methods for solving the SEP (see [50, 58] and other references therein).
If \(\mathcal {H}_2=\mathcal {H}_3\) and \(\mathcal {F}_2=I \) (I is the identity operator), (1.3) reduces to the split feasibility problem (SFP) proposed by Censor et al. [12] and defined as follows:
where \(\mathcal {F}_1:\mathcal {H}_1\rightarrow \mathcal {H}_2\) is a bounded linear operator. One of the most common method for solving (1.3) is the CQ projection method proposed and studied by Bryne et al. [9]. They defined it as follows:
where \(\eta _n\in \Big (\epsilon , \frac{2}{\lambda _{\mathcal {F}_1}+\lambda _{\mathcal {F}_2}}-\epsilon \Big ),\) and \(\lambda _{\mathcal {F}_1}\) and \(\lambda _{\mathcal {F}_2}\) are the matrix operator norms \(\Vert \mathcal {F}_1\Vert \) and \(\Vert \mathcal {F}_2\Vert ,\) respectively. Note that the step size \(\eta _n\) in Algorithm (1.4) is dependent on the operator norms, which are difficult and sometimes impossible to compute. Several authors have studied several effective methods for solving SFP (see [50] and other references therein).
Another problem of interest in this study is the fixed point problem (FPP), which is formulated as follows:
where \({T}:\mathcal {H}\rightarrow \mathcal {H}\) is a nonlinear mapping. We denote the set of fixed points of T by F(T). Several problems in sciences and engineering can be formulated as the problem of finding solutions of FPP of nonlinear mappings.
If \(\mathcal {C}\) and \(\mathcal {Q}\) are the sets of fixed points of some nonlinear operators, the SEP (1.3) becomes the split equality common fixed point problem (SECFPP) which is defined as
where \(F({T}_1)\ne \emptyset \) and \(F({T}_2)\ne \emptyset \) are the sets of fixed points of \({T}_1\) and \({T}_2,\) respectively, \({T}_1:\mathcal {H}_1\rightarrow \mathcal {H}_1\) and \({T}_2:\mathcal {H}_2\rightarrow \mathcal {H}_2\) are nonlinear mappings and \(\mathcal {F}_1:\mathcal {H}_1\rightarrow \mathcal {H}_3,~\mathcal {F}_2:\mathcal {H}_2\rightarrow \mathcal {H}_3\) are bounded linear operators.
If \(\mathcal {H}_2=\mathcal {H}_3\) and \(\mathcal {F}_2=I,\) then the SECFPP (1.5) reduces to the following split common fixed point problem (SCFPP) introduced by Censor et al. [11]
Several authors have studied and proposed effective methods for solving SCFPP (see [49] and other references therein).
The SECFPP was first studied by Moudafi et al. [37]. They introduced the following simultaneous iterative method for solving the SECFPP
where \(\eta _n\in \Big (\epsilon , \frac{2}{\lambda _{\mathcal {F}_1}+\lambda _{\mathcal {F}_2}}-\epsilon \Big ), \lambda _{\mathcal {F}_1}\) and \(\lambda _{\mathcal {F}_2}\) are the spectral radius of \(\mathcal {F}_1^*\mathcal {F}_1\) and \(\mathcal {F}_2^*\mathcal {F}_2,\) respectively, and \({T}_1\) and \({T}_2\) are firmly quasi-nonexpansive mappings. We also observe that the step size of Algorithm (1.6) depends on the operator norms. Hence to implement Algorithm (1.6), one has to compute the operator norms of \(\mathcal {F}_1\) and \(\mathcal {F}_2\) which are difficult to compute. Several authors have studied and proposed modifications of Algorithm (1.6) for better implementation (see [36, 37, 64] and other references therein).
Recently, Lopéz et al. [32] studied and proposed a method for estimating the step size which does not require prior knowledge of the operator norms for solving the SFP. Dong et al. [17] and J. Zhao [63] also proposed new choices of step size which do not require prior knowledge of the operator norm for solving SECFPP. Zhao [63] studied the SEP and presented the following step size which guarantees convergence of the iterative method without prior knowledge of the operator norm of \(\mathcal {F}_1\) and \(\mathcal {F}_2\)
The main purpose of this work is to find a common element of split equalities of the VIP, EP and common fixed point of nonexpansive semigroups. Several algorithms have been proposed for approximating the common solution of VIP, EP and FPP due to the applications it has on mathematical models whose constraints can be expressed as VIP, EP and FPP. Particularly, finding common solution problems has application in signal processing, network resource allocation, image recovery, among others (see [26, 33, 34] and other references therein).
Recently, Latif and Eslamian [31] studied and introduced a new algorithm for finding a common element of split equalities of EP, monotone VIP with Lipschitz operator and fixed point problem of nonexpansive semigroups satisfying the uniformly asymptotically regularity (u.a.r) condition in Hilbert spaces. The authors obtained strong convergence result for the proposed algorithm. However, their proposed algorithm has certain drawbacks. For instance, their method requires computing two projections each per iteration onto \(\mathcal {C}\) and \(\mathcal {Q},\) which makes it computationally expensive to implement. Moreover, the associated cost operators for the VIP are required to be monotone and Lipschitz continuous and the step size of the algorithm depends on the Lipschitz constants of these operators. In addition, the authors needed to impose the uniformly asymptotically regularity condition on the nonexpansive semigroups to obtain their result. All of these drawbacks limit the scope of application of their proposed method.
The inertial technique has been employed by several authors to increase the convergence rate of iterative methods. Polyak [45] studied the convergence of the following inertial extrapolation algorithm
where \(\alpha _1\) and \(\alpha _2\) are two real numbers. Recently, there has been an increased interest in studying inertial type algorithm (see [2, 5, 6, 23, 28, 59] and other references therein).
Motivated by the above results in the literature and other related results in this direction, we propose and study an inertial Tseng’s extragradient algorithm for the SEP for finding a common element of solution of the EP, VIP and common fixed point of nonexpansive semigroups with the following features:
-
(i)
Different from other existing methods for finding a common element of the solution of the EP, VIP and fixed point problem of nonexpansive semigroups, our method only requires that the underlying operator for the VIP be pseudomonotone, uniformly continuous and without the weak sequential continuity condition often used in the literature. Also, we do not need to assume the u.a.r condition employed by authors in the literature to obtain our strong convergence result.
-
(ii)
Different from other existing methods in the literature for solving non-Lipschitz VIP, our method does not require any linesearch technique but rather uses an easily implementable self-adaptive step size technique that generates non-monotonic sequence of step sizes. Also, our method only requires one projection each per iteration onto the feasible sets \(\mathcal {C}\) and \(\mathcal {Q}.\)
-
(iii)
Our method employs the inertial extrapolation technique to increase the rate of convergence (see [4,5,6] and other references therein).
-
(iv)
The proof of our strong convergence result does not rely on the usual “two cases approach" widely used in many papers to prove strong convergence results.
Finally, we provide some numerical experiments for our proposed method in comparison with the related method in the literature to show the applicability of our proposed method.
The rest of the paper is organized as follows: In Section 2 we present some definitions and lemmas needed to obtain the strong convergence result. In Section 3, we present our proposed method and discuss some of its important features. In Section 4, the convergence of our method is investigated and in Section 5, we present some numerical experiments of our method in comparison with a related method in the literature. We conclude in Section 6.
2 Preliminaries
In this section, we recall some lemmas, results and definitions which will be required in subsequent sections to obtain our strong convergence result. Let \(\mathcal {H}\) be a real Hilbert space with inner product \(\langle \cdot ~ \cdot \rangle \), and associated norm \(||\cdot ||\) defined by \(||x||=\sqrt{\langle x, x\rangle },\) \(\forall x\in \mathcal {H}\). We denote the strong and weak convergence by “\(\rightarrow \)” and “\(\rightharpoonup \)”, respectively. Also, we denote the set of weak limits of \(\{x_n\}\) by \(w_{\omega }(x_n),\) that is
Definition 2.1
Let \(T:\mathcal {H}\rightarrow \mathcal {H}\) be a mapping. Then, T is said to be
-
(i)
L-Lipschitz continuous, if there exists a constant \(L>0\) such that
$$\begin{aligned} \Vert Tx-Ty\Vert \le L\Vert x-y\Vert ,~~ \forall x,y\in \mathcal {H}; \end{aligned}$$if \(L\in [0,1),\) then T is called a contraction;
-
(ii)
Uniformly continuous, if for every \(\epsilon >0,\) there exists \(\delta =\delta (\epsilon )>0,\) such that
$$\Vert Tx-Ty\Vert<\epsilon \quad \text {whenever}\quad \Vert x-y\Vert <\delta ,~~\forall x,y\in \mathcal {H};$$ -
(iii)
Sequentially weakly continuous, if for each sequence \(\{x_n\},\) we have \(x_n\rightharpoonup x\in \mathcal {H}\) implies that \(Tx_n\rightharpoonup Tx\in \mathcal {H};\)
-
(iv)
Nonexpansive if T is 1-Lipschitz continuous;
-
(v)
Firmly nonexpansive if
$$\Vert Tx-Ty\Vert ^2\le \Vert x-y\Vert ^2-\Vert (I-T)x-(I-T)y\Vert ^2,~~\forall x,y\in \mathcal {H}.$$
More information on firmly nonexpansive mappings can be found, for example, in [21, Section 11]. Observe that uniform continuity is a weaker notion than Lipschitz continuity.
Definition 2.2
A one-parameter family mapping \(\mathcal {T}=\{T(s)\,:\,0\le s<+\infty \}\) from \(\mathcal {H}_1\) into itself is said to be a nonexpansive semigroup if it satisfies the following conditions:
-
(i)
\(T(0)x=x,\) \(\forall x\in \mathcal {H}_1;\)
-
(ii)
\(T(s+u)=T(s)T(u)\) for all \(s,u\ge 0;\)
-
(iii)
For each \(x\in \mathcal {H}_1,\) the mapping T(s)x is continuous;
-
(iv)
\(\Vert T(s)x-T(s)y\Vert \le \Vert x-y\Vert \) for all \(x,y\in \mathcal {H}_1\) and \(s\ge 0.\)
We denote the common fixed point set of the semigroup \(\mathcal {T}\) by \(F(\mathcal {T})=\{x\in \mathcal {C}\,:\,T(s)x=x,~\forall s\ge 0\}.\) It is well known that \(F(\mathcal {T})\) is closed and convex [8].
Lemma 2.3
[48, 55] Let \(\mathcal {C}\) be a nonempty bounded closed and convex subset of a real Hilbert space \(\mathcal {H}.\) Let \(\mathcal {T}=\{T(s)\,:\,s\ge 0\}\) from \(\mathcal {C}\) be a nonexpansive semigroup on \(\mathcal {C}\). Then for all \(h\ge 0,\)
Lemma 2.4
[55] Let \(\mathcal {C}\) be a nonempty bounded closed and convex subset of a real Hilbert space \(\mathcal {H}.\) Let \(\{x_n\}\) be a sequence and let \(\mathcal {T}=\{T(s)\,:\,s\ge 0\}\) from \(\mathcal {C}\) be a nonexpansive semigroup on \(\mathcal {C},\) if the following conditions are satisfied
-
(i)
\(x_n\rightharpoonup x;\)
-
(ii)
\(\limsup _{s \rightarrow \infty }\limsup _{n \rightarrow \infty }\Vert T(s)x_n-x_n\Vert =0,\)
then, \(x\in F(\mathcal {T}).\)
It is well known that if D is a convex subset of \(\mathcal {H},\) then \(T:D\rightarrow \mathcal {H}\) is uniformly continuous if and only if, for every \(\epsilon >0,\) there exists a constant \(M<+\infty \) such that
For the proof of (2.1), see [61, Theorem 1].
Lemma 2.5
[38] Let \(\mathcal {H}\) be a real Hilbert space, then the following assertions hold:
-
(1)
\(2\langle x, y \rangle =\Vert x\Vert ^2+\Vert y\Vert ^2-\Vert x-y\Vert ^2=\Vert x+y\Vert ^2-\Vert x\Vert ^2-\Vert y\Vert ^2,~\forall x,y \in \mathcal {H};\)
-
(2)
\(\Vert \alpha x+(1-\alpha )y\Vert ^2 = \alpha \Vert x\Vert ^2+(1-\alpha )\Vert y\Vert ^2-\alpha (1-\alpha )\Vert x-y\Vert ^2,~\forall x,y \in \mathcal {H},~ \alpha \in \mathbb {R};\)
-
(3)
\(\Vert x+y\Vert ^2 \le \Vert x\Vert ^2+2\langle y, x+y \rangle , ~\forall x,y \in \mathcal {H}\).
Lemma 2.6
[16] Assume that \(A:\mathcal {H} \rightarrow \mathcal {H} \) is a continuous and pseudomonotone operator. Then, x is a solution of (1.1) if and only if \(\langle Ay,y -x \rangle \ge 0,~ \forall y\in \mathcal {C}.\)
Lemma 2.7
[35] Let \(\mathcal {H}\) be a real Hilbert space and \(\mathcal {C}\) be a nonempty closed and convex subset of \(\mathcal {H}\). If the mapping \(h:[0,1] \rightarrow \mathcal {H}\) defined as \(h(t):=A(tx+(1-t)y)\) is continuous for all \(x,y \in \mathcal {C}\) (i.e. h is hemicontinuous), then \(M(A,\mathcal {C}):=\{x \in \mathcal {C}\,:\, \big <Ay,y-x \big > \ge 0,~ \forall y\in \mathcal {C}\}\subset VI(\mathcal {C}, A).\) Moreover, if A is pseudo-monotone, then \(VI(\mathcal {C}, A)\) is closed, convex and \(M(\mathcal {C}, A)=VI(\mathcal {C}, A)\).
Recall that for a nonempty, closed and convex subset \(\mathcal {C}\) of \(\mathcal {H}\), the metric projection denoted by \(P_\mathcal {C}\), is a map defined on \(\mathcal {H}\) onto \(\mathcal {C}\) which assigns to each \(x\in \mathcal {H}\), the unique point in \(\mathcal {C}\), denoted by \(P_\mathcal {C} x\) such that
Lemma 2.8
[21] Let \(\mathcal {C}\) be a closed and convex subset of a real Hilbert space \(\mathcal {H}\) and \(x,y\in \mathcal {H}.\) Then
-
(i)
\(\Vert P_{\mathcal {C}}x-P_{\mathcal {C}}y\Vert ^2\le \left\langle P_{\mathcal {C}}x-P_{\mathcal {C}}y,~x-y\right\rangle ;\)
-
(ii)
\(\Vert P_{\mathcal {C}}x-y\Vert ^2\le \Vert x-y\Vert ^2-\Vert x-P_{\mathcal {C}}x\Vert ^2.\)
Assumption 2.9
[7] Let \(\Phi :\mathcal {C}\times \mathcal {C}\rightarrow \mathbb {R}\) be a bifunction satisfying the following assumptions:
-
1)
\( \Phi (x,x)=0,~\forall x\in \mathcal {C};\)
-
2)
\(\Phi \) is monotone, i.e., \(\Phi (x,y)+\Phi (y,x)\le 0, ~\forall x\in \mathcal {C};\)
-
3)
For each \(x,y,z\in \mathcal {C}, ~\limsup _{t\rightarrow 0}\Phi (tz+(1-t)x,y)\le \Phi (x,y);\)
-
4)
For each \(x\in \mathcal {C}, ~y\rightarrow \Phi (x,y)\) is convex and lower semi continuous.
Lemma 2.10
[15] Let \(\Phi :\mathcal {C}\times \mathcal {C}\rightarrow \mathbb {R}\) be a bifunction satisfying Assumption 2.9. For any \(r>0\) and \(x\in \mathcal {H},\) define a mapping \(U^\Phi _r:\mathcal {H}\rightarrow \mathcal {C}\) as follows
Then, we have the following
-
(1)
\(U^{\Phi }_r\) is nonempty and single valued;
-
(2)
\(U^{\Phi }_r\) is firmly nonexpansive;
-
(3)
\(F(U^{\Phi }_r)=EP(\Phi )\) is closed and convex.
Definition 2.11
Assume that \(T:\mathcal {H}\rightarrow \mathcal {H}\) is a nonlinear operator with \(F(T)\ne \emptyset .\) Then \(I-T\) is said to be demiclosed at zero if for any \(\{x_n\}\) in \(\mathcal {H},\) the following implication holds:
Lemma 2.12
[54] Suppose \(\{\lambda _n\}\) and \(\{\theta _n\}\) are two nonnegative real sequences such that
If \(\sum _{n=1}^{\infty }\phi _n<\infty ,\) then \(\lim _{n\rightarrow \infty }\lambda _n\) exists.
Lemma 2.13
[3] Let \(\{a_n\}\) be a sequence of non-negative real numbers, \(\{\gamma _n\}\) be a sequence of real numbers in (0, 1) with conditions \(\sum _{n=1}^{\infty }\gamma _n=\infty \) and \(\{d_n\}\) be a sequence of real numbers. Assume that
If \(\limsup _{k \rightarrow \infty }d_{n_k}\le 0\) for every subsequence \(\{a_{n_k}\}\) of \(\{a_n\}\) satisfying the condition\(\liminf _{k \rightarrow \infty }(a_{{n_k}+1}-a_{n_k})\ge 0,\) then \(\lim _{n \rightarrow \infty }a_n=0.\)
Lemma 2.14
[44] Each Hilbert space H satisfies the Opial condition, that is, for any sequence \(\{x_n\}\) with \(x_n\rightharpoonup x,\) the inequality \(\liminf _{n\rightarrow \infty }||x_n-x||< \liminf _{n\rightarrow \infty }||x_n-y||\) holds for every \(y\in H\) with \(y\ne x.\)
3 Proposed Method
In this section, we present our proposed method and discuss its features. We begin with the following assumptions under which our strong convergence result is obtained.
Assumption 3.1
Suppose that the following conditions hold:
-
(a)
The feasible sets \(\mathcal {C}\) and \(\mathcal {Q}\) are nonempty, closed and convex subsets of the real Hilbert spaces \(\mathcal {H}_1\) and \(\mathcal {H}_2\), respectively.
-
(b)
\(A:\mathcal {H}_1 \rightarrow \mathcal {H}_1\) and \(B:\mathcal {H}_2 \rightarrow \mathcal {H}_2\) are pseudomonotone and uniformly continuous.
-
(c)
The mapping \(A:\mathcal {H}_1\rightarrow \mathcal {H}_1\) and \(B:\mathcal {H}_2 \rightarrow \mathcal {H}_2\) satisfies the following property: whenever \(\{x_n\}\subset \mathcal {C}, ~x_n\rightharpoonup x^*,\) one has \(\Vert Ax^*\Vert \le \liminf _{n\rightarrow \infty }\Vert Ax_n\Vert \) and whenever \(\{x_n\}\subset \mathcal {Q}, ~x_n\rightharpoonup x^*,\) one has \(\Vert Bx^*\Vert \le \liminf _{n\rightarrow \infty }\Vert Bx_n\Vert ,\) respectively.
-
(d)
\(\mathcal {F}_1:\mathcal {H}_1 \rightarrow \mathcal {H}_3\) and \(\mathcal {F}_2:\mathcal {H}_2 \rightarrow \mathcal {H}_3\) are bounded linear operators.
-
(e)
\(\Phi _1:\mathcal {C}\times \mathcal {C}\rightarrow \mathbb {R},\hspace{0.1cm} \Phi _2:\mathcal {Q}\times \mathcal {Q}\rightarrow \mathbb {R}\) are bifunctions satisfying Assumption 2.9 and \(\Phi _2\) is upper semi continuous in the first argument.
-
(f)
\(\mathcal {T}_a =\{T_1(s)\,:\,0\le s<\infty \}\) and \(\mathcal {T}_b=\{T_2(u)\,:\,0\le u<\infty \}\) are one-parameter nonexpansive semigroups on \(\mathcal {H}_1\) and \(\mathcal {H}_2,\) respectively.
-
(g)
The solution set \(\Gamma =\{x\in EP(\Phi _1)\cap VI( \mathcal {C}, A)\cap F(\mathcal {T}_a),~y\in EP(\Phi _2)\cap VI(\mathcal {Q}, B)\cap F(\mathcal {T}_b)\,:\,\mathcal {F}_1x=\mathcal {F}_2y\} \ne \emptyset .\)
-
(h)
\(\{\alpha _n\}\subset (0,1),\) \( \sum _{n=1}^{\infty }\alpha _n=+\infty ,~\lim _{n\rightarrow \infty }\alpha _n=0,~ 0<\lim \inf _{n\rightarrow \infty }\beta _n\le \limsup _{n\rightarrow \infty }\beta _n<1,~0<\liminf _{n\rightarrow \infty }\gamma _n\le \limsup _{n\rightarrow \infty }\gamma _n<1.\)
-
(i)
Let \(\{\epsilon _n\}\) and \(\{\zeta _n\}\) be positive sequences such that \(\lim _{n\rightarrow \infty }\frac{\epsilon _n}{\alpha _n}=0\) and \(\lim _{n\rightarrow \infty }\frac{\zeta _n}{\alpha _n}=0,\) respectively.
-
(j)
Let \(\{\sigma _n\}\) and \(\{\mu _n\}\) be nonnegative sequences such that \(\sum _{n=1}^{\infty }\sigma _n<+\infty \) and \(\sum _{n=1}^{\infty }\mu _n<+\infty ,\) respectively, \( \{t_{n,1}\},\{t_{n,2}\}\subset (0,+\infty ), \liminf r_{n,1}>0,\liminf r_{n,2}>0\).
Algorithm 3.2
Step 0: Choose sequences \(\{\beta _n\}^{\infty }_{n=1}, \{\gamma _n\}^{\infty }_{n=1},\{\theta _n\}^{\infty }_{n=1}\) and \(\{\tau _n\}^{\infty }_{n=1}\) such that the conditions from Assumption 3.1 (h)–(i) hold. Select an initial point \((x_0,y_0)\in \mathcal {H}_1\times \mathcal {H}_2,\) let \(\eta \ge 0,\) \( ~a_i \in (0,1), i=1,2,\) \(\lambda _1>0\), \(\rho _1>0\), \(\theta>0,~ \tau >0\) and set \(n:=1.\)
Step 1: Given the iterates \(x_{n-1},y_{n-1}\) and \(x_n,y_n\) for each \(n \ge 1,\) choose \(\theta _n\) such that \(0\le \theta _n \le \bar{\theta }_n\) and \(\tau _n\) such that \(0\le \tau _n \le \bar{\tau }_n\), where
Step 2: Compute
and
Step 3: Compute
and
Step 4: Compute
Step 5 Compute
Step 6: Compute
and
where the step size \(\eta _n\) is chosen such that for small enough \(\epsilon >0,\)
if \(\mathcal {F}_1w_n\ne \mathcal {F}_2\varphi _n\); otherwise, \(\eta _n=\eta .\)
Set \(n:=n+1\) and go back to Step 1.
Remark 3.3
The step sizes generated in (3.2) and (3.4) are allowed to increase per iteration. This reduces their dependence on the initial step sizes. When n is large enough the step size may not increase. We assume that Algorithm 3.2 does not terminate in a finite number of iterations.
Remark 3.4
By conditions (h) and (i), from (3.1) we observe that
Similarly, from (3.3) we have
Remark 3.5
We note that condition (c) of Assumption 3.1 is weaker than the sequentially weakly continuity condition.
We present an example which satisfies condition (c) of Assumption 3.1.
Example 3.6
Let \(A:\ell _2(\mathbb {R})\rightarrow \ell _2(\mathbb {R})\) be an operator defined by
Suppose that \(\{x_n\}\subset \ell _2(\mathbb {R})\) such that \(x_n\rightharpoonup x^*.\) Then, by the weakly lower semi-continuity of the norm we obtain
Thus,
Hence, A satisfies condition (c) of Assumption 3.1.
Remark 3.7
Since the sequences of step sizes generated by the algorithm in (3.2) and (3.4) are well defined and the limits \(\lim _{n\rightarrow \infty }\lambda _n\) and \(\lim _{n\rightarrow \infty }\rho _n\) exist (see Lemma 4.1). Then, the limit
Thus, there exists \(n_{0_1}> 0\) such that for all \(n> n_{0_1},\) we have \(\Big (1-\frac{\lambda _n^2a_1^2}{\lambda _{n+1}^2}\Big )>0.\)
Similarly, we have that
and there exists \(n_{0_2}> 0\) such that for all \(n> n_{0_2},\) we have \(\Big (1-\frac{\rho _n^2a_2^2}{\rho _{n+1}^2}\Big )>0.\) Now, we set \(n_0=\max \{n_{0_1}, n_{0_2}\}.\)
Remark 3.8
From the definition of \(\eta _n,\) that is,
we have
Expanding the last inequality, we have
4 Convergence Analysis
Lemma 4.1
Let \(\{\lambda _n\}\) and \(\{\rho _n\}\) be sequences generated by Algorithm 3.2. Then, we have \(\lim _{n\rightarrow \infty }\lambda _n=\lambda ,\) where \(\lambda \in [\min \{\frac{a_1}{K_1},\lambda _1\},\lambda _1+b_1], b_1=\sum _{n=1}^{\infty }\sigma _n\) for some \(K_1>0\) and\(\lim _{n\rightarrow \infty }\rho _n=\rho ,\) where \(\rho \in [\min \{\frac{a_2}{K_2},\rho _1\},\rho _1+b_2],b_2=\sum _{n=1}^{\infty }\mu _n\) for some \(K_2>0.\)
Proof
Since A is uniformly continuous, we obtain from (2.1) that for any given \(\epsilon >0,\) there exists a constant \(M<+\infty \) such that \(\Vert Au_n-A\phi _n\Vert \le M\Vert u_n-\phi _n\Vert +\epsilon .\) Thus, when \(Au_n-A\phi _n\ne 0\) for all \(n\ge 1\) we have
where \(\epsilon =\min \{\epsilon _1\Vert u_n-\phi _n\Vert \,:\, n\in \mathbb {N}\}\) for some \(\epsilon _1>0\) and \(K_1=M+\epsilon _1.\) Hence, from the definition of \(\lambda _{n+1},\) the sequence \(\{\lambda _n\}\) is bounded below by \(\min \{\frac{a_1}{K_1},\lambda _1\}\) and above by \(\lambda _1 + b_1.\) By Lemma 2.12, it follows that \(\lim _{n\rightarrow \infty }\lambda _n\) denoted by \(\lambda =\lim _{n\rightarrow \infty }\lambda _n\) exists. Clearly, we have \(\lambda \in [\min \{\frac{a_1}{K_1},\lambda _1\},\lambda _1+b_1].\)
Similarly, we have \(\lim _{n\rightarrow \infty }\rho _n=\rho ,\) and \(\rho \in \text{ min }\{\frac{a_2}{K_2},\rho _1\},\rho _1+b_2.\) \(\square \)
Lemma 4.2
Let \(\big \{(x_n,y_n)\big \}\) be a sequence generated by Algorithm 3.2 under Assumption 3.1. Then
Proof
Let \((x^*,y^*)\in \Gamma .\) Then, by applying Lemma 2.5, we have
Similarly, we have
Adding (4.1) and (4.2), we have
By (3.8) and the fact that \(\mathcal {F}_1x^*=\mathcal {F}_2y^*,\) we have
which is the desired result.\(\square \)
Lemma 4.3
Let \(\big \{(x_n,y_n)\big \}\) be a sequence generated by Algorithm 3.2 under Assumption 3.1. Then
and
Proof
Let \((x^*,y^*)\in \Gamma .\) Since \(U^{\Phi _1}_{r_{n,1}}\) is firmly nonexpansive, it follows from Lemma 2.8 that
Similarly, we have
From (3.2), we obtain
which implies that
Similarly, we have
From the definition of \(v_n\) in Step 3 and Lemma 2.5, we have
Since \(u_n=P_{\mathcal {C}}(\phi _n-\lambda _nA\phi _n)\) and \(x^*\in \mathcal {C},\) we obtain from the characteristic property of \(P_{\mathcal {C}}\) that
This implies that
Also since \(u_n\in \mathcal {C}\) and \(x^*\in \Gamma \) we have
Applying (4.4), (4.5), (4.7) and (4.8) in (4.6), we obtain
Following the same line of argument, we have
which completes the proof. \(\square \)
Lemma 4.4
Let \(\{(x_n,y_n)\}\) be a sequence generated by Algorithm 3.2 satisfying Assumption 3.1. Then \(\{(x_n,y_n)\}\) is bounded.
Proof
Let \(x^*\in \Gamma . \) From the definition of \(w_n\) and Lemma 2.5, we have
By (3.5), we have
Thus, there exists a constant \(M_1>0\) such that \((1-\alpha _n)\frac{\theta _n}{\alpha _n}\Vert x_n-x_{n-1}\Vert +\Vert x^*\Vert \le M_1\) for all \(n\in \mathbb {N}.\) Thus, from (4.11) it follows that
Consequently, we have
Following similar procedure, we have
Adding (4.12) and (4.13), we obtain
where \(c_n=2(M_1\Vert x_n-x^*\Vert +M_2\Vert y_n-y^*\Vert )+M_1^2+M^2_2.\) From STEP 3, and by applying Lemma 2.5, (4.9) together with Remark 3.6, we have
Similarly, from STEP 5, and by applying Lemma 2.5, (4.10) together with Remark 3.6, we have
From (4.3), (4.14), (4.16) and (4.18), we have
Thus, \(\big \{(x_n,y_n)\big \}\) is bounded. Consequently, \(\{z_n\},\{v_n\},\{k_n\}\) and \(\{b_n\}\) are also bounded. \(\square \)
Lemma 4.5
Let \(\big \{(x_n,y_n)\big \}\) be a sequence generated by Algorithm 3.2 under Assumption 3.1. Then,
where \(d_n=[2(1-\alpha _n)\Vert x_n-x^*\Vert \frac{\theta _n}{\alpha _n}\Vert x_n-x_{n-1}\Vert +\theta _n\Vert x_n-x_{n-1}\Vert \cdot \frac{\theta _n}{\alpha _n}\Vert x_n-x_{n-1}\Vert +2\Vert x^*\Vert \Vert w_n-x_{n+1}\Vert +2\langle x^*, x^*-x_{n+1}\rangle ]+[2(1-\alpha _n)\Vert y_n-y^*\Vert \frac{\tau _n}{\alpha _n}\Vert y_n-y_{n-1}\Vert +\tau _n\Vert y_n-y_{n-1}\Vert \cdot \frac{\tau _n}{\alpha _n}\Vert y_n-y_{n-1}\Vert +2\Vert y^*\Vert \Vert \varphi _n-y_{n+1}\Vert +2\langle y^*, y^*-y_{n+1}\rangle ].\)
Proof
Let \((x^*,y^*)\in \Gamma . \) From Lemma 2.5 and the definition of \(w_n,\) we have
Following the same line of argument, we have
Adding (4.19) and (4.20) we have
From (4.3), (4.15), (4.17) and (4.21), we have
which is the required result. \(\square \)
Now we are in a position to state the main result of this work.
Theorem 4.6
Let \(\{(x_n,y_n)\}\) be a sequence generated by Algorithm 3.2 such that Assumption 3.1 holds. Then, the sequence \(\{(x_n,y_n)\}\) converges strongly to \( (\hat{x},\hat{y})=P_{\Gamma }(0_{\mathcal {H}_1},0_{\mathcal {H}_2})\in \Gamma .\)
Proof
Let \( (\hat{x},\hat{y})=P_{\Gamma }(0_{\mathcal {H}_1},0_{\mathcal {H}_2})\in \Gamma .\) Then, it follows from Lemma 4.5 that
where \(\hat{d}_n= [2(1-\alpha _n)\Vert x_n-\hat{x}\Vert \frac{\theta _n}{\alpha _n}\Vert x_n-x_{n-1}\Vert +\theta _n\Vert x_n-x_{n-1}\Vert \cdot \frac{\theta _n}{\alpha _n}\Vert x_n-x_{n-1}\Vert +2\Vert \hat{x}\Vert \Vert w_n-x_{n+1}\Vert +2\langle \hat{x}, \hat{x}-x_{n+1}\rangle ]+[2(1-\alpha _n)\Vert y_n-\hat{y}\Vert \frac{\tau _n}{\alpha _n}\Vert y_n-y_{n-1}\Vert +\tau _n\Vert y_n-y_{n-1}\Vert \cdot \frac{\tau _n}{\alpha _n}\Vert y_n-y_{n-1}\Vert +2\Vert \hat{y}\Vert \Vert \varphi _n-y_{n+1}\Vert +2\langle \hat{y}, \hat{y}-y_{n+1}\rangle ].\) Now, we claim that the sequence \(\{\Vert x_n-\hat{x}\Vert +\Vert y_n-\hat{y}\Vert \}\) converges to zero. To show this, by Lemma 2.13 it suffices to show that \(\limsup _{k\rightarrow \infty }\hat{d}_{n_k}\le 0\) for every subsequence \(\{\Vert x_{n_k}-\hat{x}\Vert +\Vert y_{n_k}-\hat{y}\Vert \}\) of \(\{\Vert x_n-\hat{x}\Vert +\Vert y_n-\hat{y}\Vert \}\) satisfying
Suppose that \(\{\Vert x_{n_k}-\hat{x}\Vert +\Vert y_{n_k}-\hat{y}\Vert \}\) is a subsequence of \(\{\Vert x_n-\hat{x}\Vert +\Vert y_n-\hat{y}\Vert \}\) such that (4.23) holds. Again, from Lemma 4.5, we obtain
From (4.23) and the condition on \(\alpha _{n_k}\) we have
From (3.6), (3.7) and the conditions on the control parameters, we have
Also, we have
which implies that
From the definition of \(z_{n_k}, k_{n_k}\) and the previous inequality we have
Also, from the definition of \(v_{n_k},b_{n_k}\) and (4.24), we have
From (4.25) and Lemma 2.3 we have
Similarly, we have
From the definition of \(x_{n_{k+1}}\) and (4.25), we have
Similarly, we have
Now, from Step 2 and by Remark 3.4, we get
Similarly, we have
From (4.28) and (4.30) we have
Similarly, from (4.29) and (4.30)
To complete the proof, we show that \(w_{\omega }\big (x_n,y_n\big )\subset \Gamma ,\) where \(w_{\omega }\big (x_n,y_n\big )\) is the set of weak limits of \(\big \{(x_n,y_n)\big \}.\) Since \(\big \{(x_n,y_n)\big \}\) is bounded we have that \(w_{\omega }\big (x_n,y_n\big )\) is nonempty. Let \((x^*,y^*)\in w_{\omega }\big (x_n,y_n\big )\) be an arbitrary element. From (4.26), (4.28) and (4.29) we have \(x^*\in w_{\omega }(x_n)\) and \(y^*\in w_{\omega }(y_n).\) Then there exists a subsequence \(\{x_{n_k}\}\) of \(\{x_n\}\) such that \(x_{n_k}\rightharpoonup x^*\) as \(k\rightarrow \infty .\) Since \(\lim _{k\rightarrow \infty }\Vert x_{n_k}-\phi _{n_k}\Vert =0,\) we have that \(\phi _{n_k}\rightharpoonup x^*\in \mathcal {C}\) as \(k\rightarrow \infty .\) From the characteristic property of \(P_{\mathcal {C}},\) we have
which implies that
Consequently, we have
Applying the fact that \(\lim _{k \rightarrow \infty }\Vert \phi _{n_k}-u_{n_k}\Vert =0\) and \(\lim _{k\rightarrow \infty }\lambda _{n_k}=\lambda >0\) to (4.33), we have
Also, we have that
Since A is uniformly continuous on \(\mathcal {H}\) and \(\lim _{k \rightarrow \infty }\Vert \phi _{n_k}-u_{n_k}\Vert ,\) we have
Let \(\{\delta _k\}\) be a sequence of positive numbers such that \(\delta _{k+1} \le \delta _k, ~ \forall k\ge 1 \text { and } \delta _k \rightarrow 0 \text { as } k \rightarrow \infty .\) Then, for each \(k\ge 1,\) we denote by \(N_k\) the smallest positive integer such that
where the existence of \(N_k\) follows from (4.36). We have that \(\{N_k\}\) is increasing since \(\{\delta _k\}\) is decreasing. Furthermore, since \(\{u_{n_k}\}\subset \mathcal {C}\) we can suppose \(Au_{N_k} \ne 0\) (otherwise, \(u_{N_k}\) is a solution) and we set for each \(k \ge 1,~~ h_{N_k}= \frac{Au_{N_k}}{\Vert Au_{N_k}\Vert ^2}.\) Then we have that \(\langle Au_{N_k}, h_{N_k} \rangle =1~~\text {for ~each}~~ k\ge 1.\) Thus, by (4.37), we have that
which implies by the pseudo-monotonicity of A that
Since \(u_{n_k}\subset C,\) the sequence \(\{u_{n_k}\}\) converges weakly to \(x^*\in \mathcal {C}.\) If \(Ax^*=0,\) then \(x^*\in VI(\mathcal {C}, A).\) On the contrary, we suppose \(Ax^*\ne 0.\) Since A satisfies condition (c), we have
Since \(\{u_{N_k}\} \subset \{u_{n_k}\},\) we obtain that
Therefore, \(\lim _{k \rightarrow \infty } \Vert \delta _kh_{N_k}\Vert =0.\) Letting \(k \rightarrow \infty \) in (4.38) gives
which implies by Lemma 2.7 that \(x^*\in VI(\mathcal {C}, A)\). By similar argument, we have that \(y^*\in VI(\mathcal {Q},B).\)
Now, to show that \(x^*\in F(\mathcal {T}_a)\) and \(y^*\in F(\mathcal {T}_b).\) On the contrary, we suppose that \(T_1(v)x^*\ne x^*\) and \(T_2(b)x^*\ne y^*\) for all \(v\ge 0\) and \(b\ge 0.\) Then, it follows from the Opial condition of Hilbert space and from (4.27) that
which is a contradiction. Thus, it follows that \(T_1(v)x^*=x^*\) for all \(v\ge 0\) which implies that \(x^*\in F(\mathcal {T}_a).\) Similarly, \(y^*\in F(\mathcal {T}_b).\)
Next, from (4.24) we have that \(\lim _{k \rightarrow \infty }\Vert \phi _{n_k}-z_{n_k}\Vert =\lim _{k \rightarrow \infty }\Vert U^{\Phi _1}_{r_{{n_k},1}}z_{n_k}-z_{n_k}\Vert =0,\) and since \(z_{n_k}\rightharpoonup x^*\) it follows from the demiclosed property of nonexpansive mappings that \(x^*\in EP(\Phi _1).\) Similarly, we have that \(y^*\in EP(\Phi _2).\) Since \(\mathcal {F}_1x^*-\mathcal {F}_2y^*\in w_{\omega }(\mathcal {F}_1w_n-\mathcal {F}_2\varphi _n),\) it follows from the weakly lower semi-continuity of the norm that
Hence, we have that \((x^*,y^*)\in \Gamma .\) Since \((x^*,y^*)\in w_{\omega }\big (x_n,y_n\big )\) was chosen arbitrarily, it follows that \(w_{\omega }\big (x_n,y_n\big )\subset \Gamma .\) To conclude, we show that
By the boundedness of \(\{(x_{n_k}, y_{n_k})\},\) it follows that there exists a subsequence \(\{(x_{n_{k_j}}, y_{n_k})\}\) of \(\{(x_{n_k}, y_{n_k})\}\) which converges weakly to some \((\bar{x}, \bar{x})\in \mathcal {H},\) and such that
From (4.39) and the fact that \( (\hat{x},\hat{y})=P_{\Gamma }(0_{\mathcal {H}_1},0_{\mathcal {H}_2})\in \Gamma \) we have
From (4.31), (4.32) and (4.40), it follows that
Thus, by (4.30) and (4.41) we have \(\limsup _{k\rightarrow \infty } \hat{d}_{n_k} \le 0.\) Now, applying Lemma 2.13 to (4.22) we have \(\{\Vert x_n-\hat{x}\Vert +\Vert y_n-\hat{y}\Vert \}\) converges to zero, which implies that \(\lim _{n\rightarrow \infty }\Vert x_n-\hat{x}\Vert =0\) and \(\lim _{n\rightarrow \infty }\Vert y_n-\hat{y}\Vert =0.\) Therefore, \((\{x_n\},\{y_n\})\) converges strongly to \((\hat{x},\hat{y})\). \(\square \)
Example 5.1: Case 1
Example 5.1: Case 2
Example 5.1: Case 3
Example 5.1: Case 4
Example 5.2: Case 1
Example 5.2: Case 2
Example 5.2: Case 3
Example 5.2: Case 4
5 Numerical Experiment
In this section, we discuss the numerical behavior of our method, (Proposed Alg.) Algorithm 3.2 in comparison with the method in Appendix A proposed by Latif and Eslamian [31] (Latif and Eslamian Alg.), which is the only related result we could find in the literature. We plot the graph of errors against the number of iterations in each case of both examples using \(\vert x_{n+1}-x_{n}\vert < 10^{-4}\) and \(\Vert x_{n+1}-x_{n}\Vert < 10^{-4}\) in Example 5.1 and Example 5.2 respectively as the stopping criterion. The numerical computations are reported in Figs. 1, 2, 3, 4, 5, 6, 7, and 8 and Tables 1 and 2 with all implementations performed using Matlab 2021 (b).
In our computation, we choose \(\theta =3.5,~\tau =2.44,~\lambda _1=1.5,~\rho _1=1.8,~a_1=0.8, a_2=0.9, ~\epsilon _n=\zeta _n=\frac{1}{(2n+1)^3},~\alpha _n=\frac{3}{2n+1},~~\beta _n=\frac{1}{4},~\gamma _n=\frac{1}{4},~ ~~\rho _n=\sigma _n=\frac{100}{(n+1)^2}, \eta =0.5, r_{n,1}=2.8, r_{n,2}=3.5, t_{n,1}=4.5, t_{n,2}=5.5, s=u=1.5.\) For Appendix A, we choose \(\alpha = 0.85, \varsigma _n=\kappa _n=\frac{1}{6},~\xi _n=\delta _n=\frac{1-\alpha _n}{2}.\)
Example 5.1
Let \(\mathcal {H}_1=\mathcal {H}_2=\mathcal {H}_3=\mathbb {R}\) the set of all real numbers with the inner product \(\langle x,y\rangle =xy,~\forall x,y\in \mathbb {R}\) and induced norm \(|\cdot |.\) For \(r_i>0,~i=1,2,\) consider \(\mathcal {C}=[-10,10]\) and \(\mathcal {Q}=[0,20].\) We define the bifunction \(\Phi _1:\mathcal {C}\times \mathcal {C}\rightarrow \mathbb {R}\) and \(\Phi _2:\mathcal {Q}\times \mathcal {Q}\rightarrow \mathbb {R}\) as follows:
and
Let \(\mathcal {F}_1x=2x\) and \(\mathcal {F}_2x=5x\) which implies that \(\mathcal {F}_1^*x=2x\) and \(\mathcal {F}_2^*x=5x.\) Next we define \(A:\mathcal {H}_1\rightarrow \mathcal {H}_1\) as \(Ax=2x\) and \(B:\mathcal {H}_2\rightarrow \mathcal {H}_2\) as \(Bx=3x.\) We define the mappings \(T_1(s):\mathbb {R}\rightarrow \mathbb {R}\) and \(T_2(u):\mathbb {R}\rightarrow \mathbb {R}\) as follows; \(T_1(s)x=10^{-s}x\) and \(T_2(u)y=10^{-2u}y.\) Clearly, we observe that \(T_1(s)\) and \(T_2(u)\) are nonexpansive semigroups.
We choose \(\mathcal {V}_1=x_0, \mathcal {V}_2=y_0\) and consider the following cases for the numerical experiments of this example.
Case 1: Take \((x_0,y_0)=(-13.5,8.0)\) and \((x_1,y_1)=(5.7,-9.1).\)
Case 2: Take \((x_0,y_0)=(15.1,7.9)\) and \((x_1,y_1)=(6.4,81.3).\)
Case 3: Take \((x_0,y_0)=(10.9,-11.8)\) and \((x_1,y_1)=(-37.2,26.8).\)
Case 4: Take \((x_0,y_0)=(-14.9,-9.8)\) and \((x_1,y_1)=(-25.2,-17.7).\)
Example 5.2
Let \(\mathcal {H}_1=\mathcal {H}_2=\mathcal {H}_3=( l _2(\mathbb {R}), \Vert \cdot \Vert _2),\) where \( l _2(\mathbb {R}):=\{x=(x_1,x_2,\ldots ,x_n,\ldots ),\) \( x_i\in \mathbb {R}\,:\,\sum _{i=1}^{\infty }|x_i|^2<+\infty \}, ||x||_2=\sqrt{(\sum _{i=1}^{\infty }|x_i|^2)}\) and \(\langle x,y \rangle = \sum _{i=1}^\infty x_iy_i\) for all \(x\in \ell _2(\mathbb {R}).\) For \(r_i>0,~i=1,2,\) we define the sets \(\mathcal {C}:=\{x\in \ell _2\,:\,\Vert x\Vert \le 1\}\) and \(\mathcal {Q}:=\{y\in \ell _2\,:\,\Vert y\Vert \le 1\}.\) Let \(\mathcal {F}_1:\mathcal {H}_1\rightarrow \mathcal {H}_2,\) \(\mathcal {F}_2:\mathcal {H}_2\rightarrow \mathcal {H}_3\) be defined by \(\mathcal {F}_1x=\frac{x}{3}\) and \(\mathcal {F}_2x=\frac{2x}{5}\) respectively which implies that \(\mathcal {F}_1^*y=\frac{y}{3}\) and \(\mathcal {F}_2^*y=\frac{2y}{5}.\) Clearly, \(\mathcal {F}_1\) and \(\mathcal {F}_2\) are bounded linear operators. We define \(\Phi _1:\mathcal {C}\times \mathcal {C}\rightarrow \mathbb {R}\) and \(\Phi _2:\mathcal {Q}\times \mathcal {Q} \rightarrow \mathbb {R}\) by \(\Phi _1(x,y) =\langle L_1x, y-x \rangle \) and \(\Phi _2(x,y) =\langle L_2x, y-x \rangle ,\) where \(L_1x =\frac{x}{3}\) and \(L_2x =\frac{x}{2}.\) Observe that \(\Phi _1\) and \(\Phi _2\) satisfy Assumption 2.9. After simple calculation and applying Lemma 2.10, we obtain
and
Let \(A:\mathcal {H}_1\rightarrow \mathcal {H}_1\) be defined by \(A(x_1,x_2,x_3,\dots )=(x_1e^{-x_1^2},0,0,\dots )\) and \(B:\mathcal {H}_2\rightarrow \mathcal {H}_2\) as \(B(x_1,x_2,x_3,\dots )=(5x_1e^{-x_1^2},0,0,\dots ).\) Clearly, we see that A and B are pseudomonotone mappings. We define the mappings \(T_1(s):\mathbb {R}\rightarrow \mathbb {R}\) and \(T_2(u):\mathbb {R}\rightarrow \mathbb {R}\) as follows; \(T_1(s)x=10^{-5s}x\) and \(T_2(u)y=10^{-3u}y.\) Clearly, we observe that \(T_1(s)\) and \(T_2(u)\) are nonexpansive semigroups.
We choose \(\mathcal {V}_1=x_0, \mathcal {V}_2=y_0\) and consider different initial values as follows:
Case 1: \(x_0 = (\frac{1}{2}, \frac{1}{4}, \frac{1}{8}, \dots ),\) \(y_0 = (\frac{1}{2}, \frac{1}{4}, \frac{1}{8}, \dots );\) \( x_1 = (\frac{1}{3}, \frac{1}{9}, \frac{1}{27},\dots ),\) \(y_1 = (\frac{1}{3}, \frac{1}{9}, \frac{1}{27},\dots );\)
Case 2: \(x_0 = (\frac{1}{2}, \frac{1}{6}, \frac{1}{18}, \dots ),\) \(y_0 = (\frac{1}{2}, \frac{1}{4},\frac{1}{8}, \dots );\) \(x_1 = (-\frac{1}{3}, \frac{1}{6}, -\frac{1}{18}, \dots ),\) \(y_1 = (-\frac{1}{3}, \frac{1}{6}, -\frac{1}{18}, \dots );\)
Case 3: \(x_0 = (\frac{3}{8}, \frac{3}{16}, \frac{3}{32}, \dots ),\) \(y_0 = (\frac{5}{9}, \frac{5}{18}, -\frac{5}{36}, \dots );\) \(x_1 = (-\frac{1}{3}, \frac{1}{9}, -\frac{1}{27}, \dots ),\) \( y_1 = (\frac{1}{2}, \frac{1}{6}, \frac{1}{12}, \dots );\)
Case 4: \(x_0 = (\frac{3}{8}, \frac{3}{16}, \frac{3}{32},\dots ),\) \(y_0 = (\frac{5}{9}, \frac{5}{18}, \frac{5}{36},\dots );\) \(x_1 = (\frac{1}{9}, \frac{1}{18}, \frac{1}{36},\dots ),\) \(y_1 = (-\frac{7}{12}, \frac{7}{24}, -\frac{7}{36}).\)
6 Conclusion
In this paper, we studied the split equalities of the VIP, EP and FPP of nonexpansive semigroups. We introduced a Tseng’s extragradient method with self-adaptive step size for approximating a common solution of the split equalities of the VIP, EP and FPP of nonexpansive semigroups in the framework of real Hilbert spaces when the cost operator of the VIP is pseudomonotone and non-Lipschitz. Without the sequential weak continuity condition on the cost operator, we obtained a strong convergence result of our proposed method. While the cost operator is non-Lipschitz, our algorithm does not involve any linesearch procedure and our strong convergence result was obtained without the usual “two cases approach" widely used in many papers. Finally, we presented some numerical experiments of our proposed method in comparison with a related method in the literature to show the applicability of our method. Our result improves, extends and generalizes several other results in the literature.
Availability of data and material
Not applicable.
References
Alakoya, T.O., Mewomo, O.T.: S-Iteration inertial subgradient extragradient method for variational inequality and fixed point problems. Optimization (2023). https://doi.org/10.1080/02331934.2023.2168482
Alakoya, T.O., Uzor, V.A., Mewomo, O.T., Yao, J.-C.,: On system of monotone variational inclusion problems with fixed-point constraint. J. Inequal. Appl. 2022 (47), 30 (2022)
Alakoya, T.O., Uzor, V.A., Mewomo, O.T. : A new projection and contraction method for solving split monotone variational inclusion, pseudomonotone variational inequality, and common fixed point problems. Comput. Appl. Math. 42(1), 33 (2023) (Paper No. 3)
Alvarez, F.: Weak convergence of a relaxed and inertial hybrid projection-proximal point algorithm for maximal monotone operators in Hilbert space. SIAM J. Optim. 14(3), 773–782 (2004)
Attouch, H., Peypouquet, J., Redont, P.: A dynamical approach to an inertial forward-backward algorithm for convex minimization. SIAM J. Optim. 24(1), 232–256 (2014)
Beck, A., Teboulle, M.: A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM J. Imaging Sci. 2(1), 183–202 (2009)
Blum, E., Oettli, W.: From optimization and variational inequalities to equilibrium problems. Math. Stud. 63, 123–145 (1994)
Browder, F.E.: Convergence of approximants to fixed points of nonexpansive nonlinear mappings in Banach spaces. Arch. Ration. Mech. Anal. 24(1), 82–89 (1967)
Byrne, C.L., Moudafi, A.: Extensions of the CQ algorithm for the split feasibility and split equality problems. J. Nonlinear Convex Anal. 18(8), 1485–1496 (2017)
Censor, Y., Gibali, A., Reich, S.: The subgradient extragradient method for solving variational inequalities in Hilbert spaces. J. Optim. Theory Appl. 148, 318–335 (2011)
Censor, Y., Segal, A.: The split common fixed point problem for directed operators. J. Convex Anal. 16, 587–600 (2009)
Censor, Y., Elfving, T.: A multiprojection algorithm using Bregman projections in a product space. Numer. Algorithms 8, 221–239 (1994)
Censor, Y., Gibali, A., Reich, S.: Strong convergence of subgradient extragradient methods for the variational inequality problem in Hilbert space. Optimization Methods and Software 26, 827–845 (2011)
Censor, Y., Gibali, A., Reich, S.: Extensions of Korpelevich’s extragradient method for the variational inequaity problem in Euclidean space. Optimization 61, 1119–1132 (2012)
Combettes, P.L., Hirstoaga, S.A.: Equilibrium programming in Hilbert spaces. J. Nonlinear Convex Anal. 6, 117–136 (2005)
Cottle, R.W., Yao, J.C.: Pseudo-monotone complementarity problems in Hilbert space. J. Optim. Theory Appl. 75, 281–295 (1992)
Dong, Q.-L., He, S.-N., Zhao, J.: Solving the split equality problem without prior knowledge of operator norms. Optimization 64, 1887–1906 (2014)
Fichera, G.: Sul problema elastostatico di Signorini con ambigue condizioni al contorno. Atti Accad. Naz. Lincei VIII. Ser. Rend. Cl. Sci. Fis. Mat. Nat. 34, 138–142 (1963)
Gibali, A., Jolaoso, L.O., Mewomo, O.T., Taiwo, A.: Fast and simple Bregman projection methods for solving variational inequalities and related problems in Banach spaces. Results Math. 75(4), 36 (2020) (Paper No. 179)
Gibali, A., Reich, S., Zalas, R.: Outer approximation methods for solving variational inequalities in Hilbert space. Optimization 66, 417–437 (2017)
Goebel, K., Reich, S.: Uniform Convexity, Hyperbolic Geometry, and Nonexpansive Mappings. Marcel Dekker, New York (1984)
Godwin, E.C., Alakoya, T.O., Mewomo, O.T., Yao, J.-C.: Relaxed inertial Tseng extragradient method for variational inequality and fixed point problems. Appl. Anal. (2022). https://doi.org/10.1080/00036811.2022.2107913
Godwin, E.C., Izuchukwu, C., Mewomo, O.T.: Image restoration using a modified relaxed inertial method for generalized split feasibility problems. Math. Methods Appl. Sci. 46(5), 5521–5544 (2023)
Godwin, E.C., Mewomo, O.T., Alakoya, T.O.: A strongly convergent algorithm for solving multiple set split equality equilibrium and fixed point problems in Banach spaces. Proc. Edinb. Math. Soc. 2(66), 475–515 (2023)
Hartman, P., Stampacchia, G.: On some non-linear elliptic diferential-functional equations. Acta Math. 115, 271–310 (1966)
Iiduka, H.: Acceleration method for convex optimization over the fixed point set of a nonexpansive mappings. Math. Prog. Series A. 149(1–2), 131–165 (2015)
Izuchukwu, C., Ogwo, G.N., Mewomo, O.T.: An inertial method for solving generalized split feasibility problems over the solution set of monotone variational inclusions. Optimization 71(3), 583–611 (2022)
Izuchukwu, C., Reich, S., Shehu, Y.: Relaxed inertial methods for solving the split monotone variational inclusion problem beyond co-coerciveness. Optimization 72, 607–646 (2023)
Kazmi, K.R., Ali, R., Furkan, M.: Common solution to a split equality monotone variational inclusion problem, a split equality generalized general variational-like inequality problem and a split equality fixed point problem. Fixed Point Theory 20(1), 211–231 (2019)
Kraikaew, R., Saejung, S.: Strong convergence of the Halpern subgradient extragradient method for solving variational inequalities in Hilbert spaces. J. Optim. Theory Appl. 163, 399–412 (2014)
Latif, A., Eslamian, M.: Split equality problem with equilibrium problem, variational inequality problem, and fixed point problem of nonexpansive semigroups. J. Nonlinear Sci. Appl. 10, 3217–3230 (2017)
López, G., Martín-Márquez, V., Wang, F.-H., Xu, H.-K.: Solving the split feasibility problem without prior knowledge of matrix norms. Inverse Prob. 27, 18 (2012)
Maingé, P.E.: A hybrid extragradient-viscosity method for monotone operators and fixed point problems. SIAM J. Control. Optim. 47, 1499–1515 (2008)
Maingé, P.E.: Projected subgradient techniques and viscosity methods for optimization with variational inequality constraints. European J. Oper. Res. 205, 501–506 (2010)
Mashreghi, J., Nasri, M.: Forcing strong convergence of Korpelevich’s method in Banach spaces with its applications in game theory. Nonlinear Anal. 72, 2086–2099 (2010)
Mewomo, O.T., Ogbuisi, F.U.: Convergence analysis of an iterative method for solving multiple-set split feasibility problems in certain Banach spaces. Quaest. Math. 41(1), 129–148 (2018)
Moudafi, A., Al-Shemas, E.: Simultaneous iterative methods for split equality problem. Trans. Math. Program. Appl. 1, 1–11 (2013)
Ogbuisi, F.U., Mewomo, O.T.: On split generalised mixed equilibrium problems and fixed-point problems with no prior knowledge of operator norm. J. Fixed Point Theory Appl. 19(3), 2109–2128 (2017)
Ogwo, G.N., Alakoya, T.O., Mewomo, O.T.: An inertial subgradient extragradient method with Armijo type step size for pseudomonotone variational inequalities with non-Lipschitz operators in Banach spaces. J. Ind. Manag. Optim. (2022). https://doi.org/10.3934/jimo.2022239
Ogwo, G.N., Izuchukwu, C., Mewomo, O.T.: Relaxed inertial methods for solving split variational inequality problems without product space formulation. Acta Math. Sci. Ser. B (Engl. Ed.) 42(5), 1701–1733 (2022)
Ogwo, G.N., Izuchukwu, C., Shehu, Y., Mewomo, O.T.: Convergence of relaxed inertial subgradient extragradient methods for quasimonotone variational inequality problems. J. Sci. Comput. 90(1), 1–35 (2021) (Paper No. 10)
Okeke, C.C., Mewomo, O.T.: On split equilibrium problem, variational inequality problem and fixed point problem for multi-valued mappings. Ann. Acad. Rom. Sci. Ser. Math. Appl. 9(2), 223–248 (2017)
Owolabi, A.O.-E., Alakoya, T.O., Taiwo, A., Mewomo, O.T.: A new inertial-projection algorithm for approximating common solution of variational inequality and fixed point problems of multivalued mappings. Numer. Algebra Control Optim. 12(2), 255–278 (2022)
Opial, Z.: Weak convergence of the sequence of successive approximation for nonexpansive mappings. Bull. Amer. Math. Soc. 73, 591–597 (1967)
Polyak, B.T.: Some methods of speeding up the convergence of iteration methods. U.S.S.R. Comput. Math. Math. Phys. 4(5), 1–17 (1964)
Qin, X., Cho, Y.J., Kang, S.M.: Convergence theorems of common elements for equilibrium problems and fixed point problems in Banach spaces. J. Comput. Appl. Math. 225, 20–30 (2009)
Reich, S., Sabach, S.: Three strong convergence theorems regarding iterative methods for solving equilibrium problems in reflexive Banach spaces. Contemporary Math. 568, 225–230 (2012)
Reich, S., Xu, H.-K.: Nonlinear ergodic theory for semigroups of Lipschitzian mappings. Comm. Appl. Nonlinear Anal. 1, 47–60 (1994)
Reich, S., Tuyen, T.M., Thuy, N.T.T., Ha, M.T.N.: A new self-adaptive algorithm for solving the split common fixed point problem with multiple output sets in Hilbert spaces. Numer. Algorithms 89, 1031–1047 (2022)
Reich, S., Tuyen, T.M.: A new approach to solving split equality problems in Hilbert spaces. Optimization 71(15), 4423–4445 (2022)
Stampacchia, G.: Variational Inequalities. In: Theory and Applications of Monotone Operators. Proceedings of the NATO Advanced Study Institute, Venice, Italy (Edizioni Odersi, Gubbio, Italy, 1968). pp. 102–192
Taiwo, A., Jolaoso, L.O., Mewomo, O.T.: Inertial-type algorithm for solving split common fixed point problems in Banach spaces. J. Sci. Comput. 86(12), 30 (2021)
Taiwo, A., Mewomo, O.T.: Inertial-viscosity-type algorithms for solving generalized equilibrium and fixed point problems in Hilbert spaces. Vietnam J. Math. 50(1), 125–149 (2022)
Tan, K.K., Xu, H.K.: Approximating fixed points of nonexpansive mappings by the Ishikawa iteration process. J. Math. Anal. Appl. 178, 301–308 (1993)
Tan, K., Xu, H.K.: The nonlinear ergodic theorem for asymptotically nonexpansive mappings in Banach spaces. Proc. Amer. Math. Soc. 114(2), 399–404 (1992)
Tian, M., Xu, G.: Inertial modified Tseng’s extragradient algorithms for solving monotone variational inequalities and fixed point problems. Nonlinear Funct. Anal. 35 (2020)
Tseng, P.: A modified forward-backward splitting method for maximal monotone mappings. SIAM J. Control. Optim. 38, 431–446 (2000)
Tuyen, T.M.: Regularization Methods for the Split Equality Problems in Hilbert Spaces. Bull. Malays. Math. Sci. Soc. 46, 44 (2023)
Uzor, V.A., Alakoya, T.O., Mewomo, O.T.: On Split Monotone Variational Inclusion Problem with Multiple Output Sets with fixed point constraints. Comput. Methods Appl. Math. (2022). https://doi.org/10.1515/cmam-2022-0199
Uzor, V.A., Alakoya, T.O., Mewomo, O.T.: Strong convergence of a self-adaptive inertial Tseng’s extragradient method for pseudomonotone variational inequalities and fixed point problems. Open Math. 20, 234–257 (2022)
Vanderbei, R.: Uniform continuity is almost Lipschitz continuity. Princeton University, Statistics and Operations Research Series (1991)
Wickramasinghe, M.U., Mewomo, O.T., Alakoya, T.O., Iyiola, O.S.: Mann-type approximation scheme for solving a new class of split inverse problems in Hilbert spaces. Appl. Anal. (2023). https://doi.org/10.1080/00036811.2023.2233977
Zhao, J.: Solving split equality fixed-point problem of quasi-nonexpansive mappings without prior knowledge of operators norms. Optimization 64, 2619–2630 (2014)
Zhao, J., Wang, S.: Viscosity approximation methods for the split equality common fixed point problem of quasi-nonexpansive operators. Acta Math. Sci. Ser. B (Engl. Ed.) 36(5), 1474–1486 (2016)
Acknowledgements
The authors thank the reviewers for the time spent and efforts made to read through the manuscript and for their constructive comments and recommendations, which have helped to improve on the quality of the article. The first author is supported by the National Research Foundation (NRF) of South Africa Incentive Funding for Rated Researchers (Grant Number 119903) and DSI-NRF Centre of Excellence in Mathematical and Statistical Sciences (CoE-MaSS), South Africa (Grant Number 2022-087-OPA). The second author acknowledges with thanks the scholarship and financial support from the University of KwaZulu-Natal (UKZN) Doctoral Scholarship. The research of the third author is wholly supported by the University of KwaZulu-Natal, Durban, South Africa Postdoctoral Fellowship. He is grateful for the funding and financial support. Opinions expressed and conclusions arrived are those of the authors and are not necessarily to be attributed to the CoE-MaSS and NRF.
Funding
Open access funding provided by University of KwaZulu-Natal. The first author is supported by the National Research Foundation (NRF) of South Africa Incentive Funding for Rated Researchers (Grant Number 119903) and DSI-NRF Centre of Excellence in Mathematical and Statistical Sciences (CoE-MaSS), South Africa (Grant Number 2022-087-OPA). The second author is funded by University of KwaZulu-Natal, Durban, South Africa Doctoral Fellowship. The third author is funded by University of KwaZulu-Natal, Durban, South Africa Postdoctoral Fellowship.
Author information
Authors and Affiliations
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.
Appendix A Algorithm 1 of Latif et al. [31]
Appendix A Algorithm 1 of Latif et al. [31]
Choose sequences \(\{\beta _{n}\}^{\infty }_{n=1}, \{\alpha _{n}\}^{\infty }_{n=1},\{\delta _{n}\}^{\infty }_{n=1}\) such that \(\beta _n+\alpha _n+\delta _n=1.\) Select initial point \(x_0\in \mathcal {H}_1, y_0\in \mathcal {H}_2,\) let \(\vartheta \ge 0.\) Set \(n:=1.\)
where the step size \(\vartheta _n\) is chosen such that for small enough \(\epsilon >0,\)
if \(\mathcal {F}_1x_n\ne \mathcal {F}_2y_n\); otherwise, \(\vartheta _n=\eta .\)
Set \(n:=n+1\) and go back to Step 1.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Mewomo, O.T., Ogwo, G.N. & Alakoya, T.O. An Inertial Iterative Algorithm for Approximating Common Solutions to Split Equalities of Some Nonlinear Optimization Problems. Acta Math Vietnam 48, 621–650 (2023). https://doi.org/10.1007/s40306-023-00521-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40306-023-00521-5
Keywords
- Split equality problems
- Equilibrium problem
- Variational inequalities
- Nonexpansive semigroup
- Inertial technique
- Self-adaptive step size