数学物理学报, 2023, 43(6): 1880-1896

Hilbert 空间中不动点问题, 变分不等式系统和分裂均衡问题迭代算法的强收敛定理

潘灵荣,1,*, 王元恒2

1浙江开放大学温岭学院 浙江温岭 317500

2浙江师范大学数学与计算机科学学院 浙江金华 321004

The Strong Convergence Theorem of Iterative Algorithms for the Fixed Point Problem, a System of Variational Inequalities, and a Split Equilibrium Problem in Hilbert Spaces

Pan Lingrong,1,*, Wang Yuanheng2

1Wenling Institute, Zhejiang Open University, Zhejiang Wenling 317500

2College of Mathematics and Computer Science, Zhejiang Normal University, Zhejiang Jinhua 321004

通讯作者: *潘灵荣,E-mail: 2671825414@qq.com

收稿日期: 2022-08-26   修回日期: 2023-05-17  

基金资助: 国家自然科学基金(12171435)
浙江省教育厅科研项目(Y202146280)

Received: 2022-08-26   Revised: 2023-05-17  

Fund supported: NSFC(12171435)
Scientific Research Fund of Zhejiang Province Education Depertment(Y202146280)

摘要

在 Hilbert 空间中, 考虑渐近非扩张映像的不动点问题, 变分不等式系统和分裂均衡问题的一个公共解. 在适当的参数限制条件下, 证明了所构造的修正粘性迭代序列强收敛到以上三类问题的一个公共元, 所得结论改进和推广了一些最近文献的相关结果.

关键词: 渐近非扩张映像; 变分不等式系统; 分裂均衡问题; 不动点问题

Abstract

In this paper, we consider a common solution of three problems in real Hilbert spaces including the fixed points problem for asymptotically nonexpansive mapping, a system of variational inequalities and the split equilibrium problem. Under some suitable conditions imposed on the sequence of parameters, we prove that the sequence generated by the modified viscosity approximation method converges strongly to a common element of the solution set of these three kinds of problems. The results obtained in this article extend and improve the corresponding results of the relevant literature.

Keywords: Asymptotically nonexpansive mapping; System of variational inequalities; Split equilibrium problem; Fixed point problem

PDF (539KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

潘灵荣, 王元恒. Hilbert 空间中不动点问题, 变分不等式系统和分裂均衡问题迭代算法的强收敛定理[J]. 数学物理学报, 2023, 43(6): 1880-1896

Pan Lingrong, Wang Yuanheng. The Strong Convergence Theorem of Iterative Algorithms for the Fixed Point Problem, a System of Variational Inequalities, and a Split Equilibrium Problem in Hilbert Spaces[J]. Acta Mathematica Scientia, 2023, 43(6): 1880-1896

1 引言

设 $H_{1}$ 和 $H_{2}$ 是具有内积 $\langle\cdot, \cdot\rangle$ 和范数 $\|\cdot\|$ 的两个实 Hilbert 空间, $C_{1}\subset H_{1}$ 和 $C_{2}\subset H_{2}$ 是两个非空闭凸子集. 令 $R, \rightarrow$ 和 $\rightharpoonup$ 分别为实数集, 强收敛和弱收敛. 设$S:C_{1}\mapsto C_{1}$ 是一个映射, 如果存在一个序列 $\{\theta_{n}\}\subset [1, \infty)$ 且 $\lim\limits_{n\rightarrow\infty}\theta_{n}=1$, 那么称 $S$ 是渐近非扩张映射, 即 $\|S^{n}x-S^{n}y\|\leq\theta_{n}\|x-y\|, \forall n\geq 1, x, y\in C_{1}.$ 记 $S$ 的不动点集为 Fix$(S)$, 显然 Fix$(S)$ 是一个非空闭凸集.

均衡问题是当前的研究热点之一, 许多关于力学、优化理论和经济学等问题都与寻找均衡问题的解相关[1-3]. 近年来, 围绕构造有效的迭代方法来寻找均衡问题解的研究成果出现较多[4-8].

设 $F_{1}:C_{1}\times C_{1}\mapsto \mathbb{R}$ 是二元函数, 均衡问题是指求一点 $x^{*}\in C_{1}$, 使得

$F_{1}(x^{*}, x)\geq0, \ \ \ \forall x\in C_{1},$

问题 (1.1) 的解集记为 $EP(F_{1}).$

设 $\varphi_{1}:C_{1}\times C_{1}\mapsto \mathbb{R}$ 是一个非线性二元函数, 广义均衡问题是指求一点 $x^{*}\in C_{1}$, 使得

$F_{1}(x^{*}, x)+\varphi_{1}(x^{*}, x)\geq0, \ \ \ \forall x\in C_{1},$

问题 (1.2) 的解集记为 $GEP(F_{1}, \varphi_{1})$, 特别地, 如果 $\varphi_{1}=0$, 那么问题 (1.2) 转化为问题 (1.1).

最近, Withun[9]考虑以下分裂广义均衡问题, 求一点 $x^{*}\in C_{1},$ 满足 (12) 式和

$y^{*}=Ax^{*}\in C_{2}, F_{2}(y^{*}, y)+\varphi_{2}(y^{*}, y)\geq0, \ \ \ \forall y\in C_{2},$

这里 $F_{1}, \varphi_{1}:C_{1}\times C_{1}\mapsto \mathbb{R}$ 和 $F_{2}, \varphi_{2}:C_{2}\times C_{2}\mapsto \mathbb{R}$ 是非线性二元函数, $A:H_{1}\mapsto H_{2}$ 是有界线性算子. 定义分裂广义均衡问题 (1.2)-(1.3) 的解集为

$SGEP(F_{1}, \varphi_{1}, F_{2}, \varphi_{2}):=\{x^{*}\in C_{1}:x^{*}\in GEP(F_{1}, \varphi_{1}), Ax^{*}\in GEP(F_{2}, \varphi_{2})\}.$

变分不等式理论是研究纯粹数学和应用数学的多个分支问题的重要工具之一, 在现实中有诸多的应用, 如博弈论、信号处理和平衡问题等, 因此对解决变分不等式问题的可行迭代方法受到了很多的关注[10-15,22].

在 2008 年, Ceng 等[16]在 Hilbert 空间 $H_{1}$ 中引入如下变分不等式系统, 即寻找 $(x^{*}, y^{*})\in C_{1}\times C_{1}$, 使得

$\begin{cases} \langle \lambda_{1}Ay^{*}+x^{*}-y^{*}, x-x^{*}\rangle\geq0, \ \ \ \forall x\in C_{1}, \\ \langle \lambda_{2}Bx^{*}+y^{*}-x^{*}, x-y^{*}\rangle\geq0, \ \ \ \forall x\in C_{1}, \\ \end{cases}$

这里 $\lambda_{1}>0, \lambda_{2}>0, A, B:C_{1}\mapsto H_{1}$ 是两个非线性映射, 问题 (1.5) 称为广义变分不等式系统.

在此基础上, 我们考虑更一般的变分不等式系统问题, 即寻找 $(u_{1}, u_{2},\cdots, u_{N})\in C_{1}\times C_{1}\times\cdots\times C_{1}$, 满足

$\begin{cases} \langle u_{N}-(I-\lambda_{N}B_{N})(au_{N}+(1-a)u_{N-1}), x-u_{N}\rangle\geq0, \ \ \ \forall x\in C_{1}, \\ \langle u_{N-1}-(I-\lambda_{N-1}B_{N-1})(au_{N}+(1-a)u_{N-2}), x-u_{N-1}\rangle\geq0, \ \ \ \forall x\in C_{1}, \\ \cdots \\ \langle u_{1}-(I-\lambda_{1}B_{1})u_{N}, x-u_{1}\rangle\geq0, \ \ \ \forall x\in C_{1}, \\ \end{cases}$

这里 $a\in[0, 1)$, 对所有 $i\in(1, 2,\cdots, N), \lambda_{i}>0, B_{i}:C_{1}\mapsto H_{1}$ 是非线性映射. 若取 $N=2$, 那么问题 (1.6) 转化为文献 [17] 中的广义变分不等式系统问题; 若取 $a=0, N=2$, 那么问题 (1.6) 转化为问题 (1.5).

本文结合渐近非扩张映射的不动点问题、分裂广义均衡问题 (1.2)-(1.3) 和变分不等式系统 (1.6), 构造了一个修正的粘性迭代序列并证明了该生成序列强收敛到这三类问题的一个公共元.

2 预备知识

设 $H$ 是实 Hilbert 空间, $C$ 是 $H$ 中的非空闭凸子集, 令 $F:C\mapsto H$ 是一个映射

(i) 如果 $\|Fx-Fy\|\leq\|x-y\|, \forall x, y\in C$, 那么称 $F$ 是非扩张的;

(ii) 如果存在一个常数 $\rho\in(0, 1)$, 使得 $\|Fx-Fy\|\leq\rho\|x-y\|, \forall x, y\in C$, 那么称 $F$ 是严格压缩映射;

(iii) 如果 $\langle Fx-Fy, x-y\rangle\geq0, \forall x, y\in C$, 那么称 $F$ 是单调的;

(iv) 如果存在一个常数 $b>0$, 使得 $\langle Fx-Fy, x-y\rangle\geq b\|Fx-Fy\|^{2}, \forall x, y\in C$, 那么称 $F$ 是 $b$-逆强单调的.

为证明本文结论, 还需要以下引理.

引理 2.1[14] 设 $H$ 是实 Hilbert 空间, $C$ 是 $H$ 中的非空闭凸子集, 如果算子 $B:C\mapsto H$ 是 $b$-逆强单调的, 那么有

$\|(I-\lambda B)x-(I-\lambda B)y\|^{2}\leq\|x-y\|^{2}+\lambda(\lambda-2b)\|Bx-By\|^{2},$

其中 $x, y\in C, \lambda>0,$ 如果 $0<\lambda<2b,$ 那么称 $I-\lambda B$ 是非扩张的.

引理 2.2[18] 设 $X$ 是 Banach 空间, $\{x_{n}\}$ 和 $\{y_{n}\}$ 是 $X$ 中的有界序列, 序列 $\{\beta_{n}\}\subset[0,1]$ 且 $0<\liminf\limits_{n\rightarrow\infty}{\beta_{n}}\leq\limsup\limits_{n\rightarrow\infty}{\beta_{n}}<1,$ 令 $x_{n+1}=\beta_{n}x_{n}+(1-\beta_{n})w_{n}, n\geq0$, 若

$\lim\limits_{n\rightarrow\infty}(\|w_{n+1}-w_{n}\|-\|x_{n+1}-x_{n}\|)\leq0,$

则 $\lim\limits_{n\rightarrow\infty}\|w_{n}-x_{n}\|=0.$

引理 2.3 假设 $H_{1}$ 是 Hilbert 空间, $C_{1}\subset H_{1}$ 是非空闭凸集, 令 $B_{i}:C_{1}\mapsto H_{1}$ 是 $b_{i}$-逆强单调映射, 这里 $i=1, 2,\cdots, N$. 对所有的 $\lambda_{i}>0$ 和 $a\in[0, 1)$, 以下结果等价

(1) $(u_{1}, u_{2},\cdots, u_{N})\in C_{1}\times C_{1}\times\cdots\times C_{1}$ 是问题 (1.6) 的解;

(2) $u_{N}$ 是映射 $G$ 的不动点, 即 $u_{N}\in$ Fix$(G), \forall x\in C_{1}$, 映射 $G:C_{1}\mapsto C_{1}$ 定义为 $G(x)=P_{C_{1}}(I-\lambda_{N}B_{N})(ax+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1})\cdots(ax+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots x,$ 且 $u_{i}=P_{C_{1}}(I-\lambda_{i}B_{i})(au_{N}+(1-a)u_{i-1}), u_{1}=P_{C_{1}}(I-\lambda_{1}B_{1})u_{N}, i= 2, 3, \cdots, N.$

$(1)\Rightarrow(2)$ 设 $(u_{1}, u_{2},\cdots, u_{N})\in C_{1}\times C_{1}\times\cdots\times C_{1}$ 是问题 (1.6) 的解, $\forall x\in C_{1}$, 则

$\begin{cases} \langle u_{N}-(I-\lambda_{N}B_{N})(au_{N}+(1-a)u_{N-1}), x-u_{N}\rangle\geq0, \nonumber\\ \langle u_{N-1}-(I-\lambda_{N-1}B_{N-1})(au_{N}+(1-a)u_{N-2}), x-u_{N-1}\rangle\geq0, \\ \cdots \\ \langle u_{1}-(I-\lambda_{1}B_{1})u_{N}, x-u_{1}\rangle\geq0. \end{cases}$

由 $P_{C_{1}}$ 的性质, 可知

$\begin{cases} u_{N}=P_{C_{1}}(I-\lambda_{N}B_{N})(au_{N}+(1-a)u_{N-1}), \nonumber\\ u_{N-1}=P_{C_{1}}(I-\lambda_{N-1}B_{N-1})(au_{N}+(1-a)u_{N-2}), \\ \cdots \\ u_{1}=P_{C_{1}}(I-\lambda_{1}B_{1})u_{N}. \end{cases}$

这说明 $u_{N}=P_{C_{1}}(I-\lambda_{N}B_{N})(au_{N}+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1})(au_{N}+(1-a) P_{C_{1}}(I-\lambda_{N-2}B_{N-2})\cdots(au_{N}+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots u_{N}=G(u_{N}),$ 即 $u_{N}\in $ Fix$(G).$

$(2)\Rightarrow(1)$ 由于 $u_{N}\in $ Fix$(G)$, 有

$G(u_{N})=P_{C_{1}}(I-\lambda_{N}B_{N})(au_{N}+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1}) (au_{N}+(1-a)\cdot \\ P_{C_{1}} (I-\lambda_{N-2}B_{N-2}) \cdots(au_{N}+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots u_{N}=u_{N},$

已知 $u_{i}=P_{C_{1}}(I-\lambda_{i}B_{i})(au_{N}+(1-a)u_{i-1}), u_{1}=P_{C_{1}}(I-\lambda_{1}B_{1})u_{N}, i= 2, 3, \cdots, N.$ 得到

$\begin{cases} u_{N}=P_{C_{1}}(I-\lambda_{N}B_{N})(au_{N}+(1-a)u_{N-1}), \nonumber\\ u_{N-1}=P_{C_{1}}(I-\lambda_{N-1}B_{N-1})(au_{N}+(1-a)u_{N-2}), \\ \cdots \\ u_{1}=P_{C_{1}}(I-\lambda_{1}B_{1})u_{N}, \ \ \ \forall x\in C_{1}, \\ \end{cases}$

那么 $(u_{1}, u_{2},\cdots, u_{N})\in C_{1}\times C_{1}\times\cdots\times C_{1}$ 是问题 (1.6) 的解. 证毕.

引理 2.4 设 $C_{1}$ 是实 Hilbert 空间 $H_{1}$ 的非空闭凸子集, $B_{i}:C_{1}\mapsto H_{1}$ 是 $b_{i}$-逆强单调映射, 这里 $i=1, 2,\cdots, N.$ 映射 $G$ 的定义见引理 2.3 且 $a\in[0, 1)$. 如果 $0<\lambda_{i}\leq 2b_{i}, i=1, 2,\cdots, N$, 那么 $G$ 是非扩张的.

因 $B_{i}$ 是 $b_{i}$-逆强单调映射, 其中 $i=1, 2,\cdots, N.$ 由引理 2.1 可知, 对于所有的 $i$, $I-\lambda_{i}B_{i}$ 是非扩张映射, 那么得到 $P_{C_{1}}(I-\lambda_{i}B_{i})$ 也是非扩张映射. 取 $x, y\in C_{1}$, 有

$\|G(x)-G(y)\|=\ \|P_{C_{1}}(I-\lambda_{N}B_{N})(ax+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1}) \\ \cdots(ax+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots x \\ -P_{C_{1}}(I-\lambda_{N}B_{N})(ay+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1}) \\ \cdots(ay+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots y\| \\ \leq\ \|(I-\lambda_{N}B_{N})(ax+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1}) \\ \cdots(ax+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots x \\ -(I-\lambda_{N}B_{N})(ay+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1}) \\ \cdots(ay+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots y\| \\ \leq \cdots\leq a\|x-y\|+(1-a)(a\|x-y\| \\ +(1-a)\|P_{C_{1}}(I-\lambda_{1}B_{1})x-P_{C_{1}}(I-\lambda_{1}B_{1})y\|) \\ \leq\ a\|x-y\|+(1-a)\|x-y\|=\|x-y\|.$

所以, $G$ 是非扩张的.

为了解决广义均衡问题, 令二元函数 $F_{1}, \varphi_{1}:C_{1}\times C_{1}\mapsto R$ 满足下列条件

(A1) 对于所有的 $x\in C_{1}, F_{1}(x, x)=0;$

(A2) $F_{1}$ 是单调的, 即 $F_{1}(x, y)+F_{1}(y, x)\leq0, \forall x, y\in C_{1};$

(A3) $F_{1}$ 是弱上半连续的, 即对于每一个 $x, y, z\in C_{1}, \lim\limits_{t\rightarrow0}F_{1}(tz+(1-t)x, y)\leq F_{1}(x, y)$;

(A4) 对于每一个 $x\in C_{1}, y\mapsto F_{1}(x, y)$ 是凸的和下半连续的;

(A5) 对于所有的 $x\in C_{1}, \varphi_{1}(x, x)\geq0;$

(A6) 对于每一个 $y\in C_{1}, x\mapsto \varphi_{1}(x, y)$ 是上半连续的;

(A7) 对于每一个 $x\in C_{1}, y\mapsto \varphi_{1}(x, y)$ 是凸的和下半连续的.

引理 2.5[9] 设 $H_{1}$ 是实 Hilbert 空间, $C_{1}$ 是 $H_{1}$ 的非空闭凸子集. 设 $F_{1}, \varphi_{1}:C_{1}\times C_{1}\mapsto R$ 是二元函数且满足条件 (A1)-(A7). 令 $\varphi_{1}$ 是单调的, 对于每一个 $r>0$ 和 $x\in H_{1}$, 定义映射 $T_{r}^{(F_{1}, \varphi_{1})}:H_{1}\mapsto C_{1}$ 如下

$T_{r}^{(F_{1}, \varphi_{1})}(x)=\{z\in C_{1}:F_{1}(z, y)+\varphi_{1}(z, y)+\frac{1}{r}\langle y-z, z-x\rangle\geq0, \forall y\in C_{1}\},$

那么以下结论成立

(1) 对于每一个 $x\in H_{1}, T_{r}^{(F_{1}, \varphi_{1})}\neq\emptyset$;

(2) $T_{r}^{(F_{1}, \varphi_{1})}$ 是单值的;

(3) $T_{r}^{(F_{1}, \varphi_{1})}$ 是定非扩张的. 即对于任何 $x, y\in H_{1}$, 有

$\|T_{r}^{(F_{1}, \varphi_{1})}x-T_{r}^{(F_{1}, \varphi_{1})}y\|^{2}\leq\langle T_{r}^{(F_{1}, \varphi_{1})}x-T_{r}^{(F_{1}, \varphi_{1})}y, x-y\rangle;$

(4) Fix$(T_{r}^{(F_{1}, \varphi_{1})})=GEP(F_{1}, \varphi_{1});$

(5) $GEP(F_{1}, \varphi_{1})$ 是紧凸集.

设 $H_{2}$ 是一个实 Hilbert 空间, $C_{2}$ 是 $H_{2}$ 的非空闭凸子集, 设 $F_{2}, \varphi_{2}:C_{2}\times C_{2}\mapsto R$ 是二元函数且满足条件 (A1)-(A7), 对于每一个 $s>0$ 和 $w\in H_{2}$, 定义映射 $T_{s}^{(F_{2}, \varphi_{2})}:H_{2}\mapsto C_{2}$ 如下

$T_{s}^{(F_{2}, \varphi_{2})}(v)=\{w\in C_{2}:F_{2}(w, d)+\varphi_{2}(w, d)+\frac{1}{s}\langle d-w, w-v\rangle\geq0, \forall d\in C_{2}\}.$

那么以下结论成立

(6) 对于每一个 $v\in H_{2}, T_{s}^{(F_{2}, \varphi_{2})}\neq\emptyset$;

(7) $T_{s}^{(F_{2}, \varphi_{2})}$ 是单值的;

(8) $T_{s}^{(F_{2}, \varphi_{2})}$ 是定非扩张的;

(9) Fix$(T_{s}^{(F_{2}, \varphi_{2})})=GEP(F_{2}, \varphi_{2})$;

(10) $GEP(F_{2}, \varphi_{2})$ 是闭凸集, 这里 $GEP(F_{2}, \varphi_{2})$ 是下列广义均衡问题的解集.

对于所有的 $y\in C_{2}$, 存在 $y^{*}\in C_{2},$ 使得

$F_{2}(y^{*}, y)+\varphi_{2}(y^{*}, y)\geq0.$

显然, $SGEP(F_{1}, \varphi_{1}, F_{2}, \varphi_{2})$ 是闭的和凸的.

引理 2.6[19] 设 $H_{1}$ 是实 Hilbert 空间, $C_{1}$ 是 $H_{1}$ 的非空闭凸子集. 设 $F_{1}, \varphi_{1}:C_{1}\times C_{1}\mapsto \mathbb{R}$ 是二元函数且满足条件 (A1)-(A7). $T_{r}^{(F_{1}, \varphi_{1})}$ 的定义见引理 2.5, 令 $x, y\in H_{1}$ 和 $r_{1}, r_{2}>0$, 有

$\|T_{r_{2}}^{(F_{1}, \varphi_{1})}y-T_{r_{1}}^{(F_{1}, \varphi_{1})}x\|\leq\|y-x\|+|\frac{r_{2}-r_{1}}{r_{2}}|\|T_{r_{2}}^{(F_{1}, \varphi_{1})}y-y\|.$

引理 2.7 设 $H$ 是 Hilbert 空间, $C$ 是 $H$ 上的非空闭凸子集, 定义 $P_{C}$ 是 $H$ 到 $C$ 上的投影, 众所周知, $P_{C}$ 是非扩张的且满足以下不等式

(1) $\|P_{C}x-P_{C}y\|^{2}\leq\langle x-y, P_{C}x-P_{C}y\rangle, \ \ \ \forall x, y\in H$;

(2) $\|x-y\|^{2}\geq\|x-P_{C}x\|^{2}+\|y-P_{C}x\|^{2}, \ \ \ \forall x\in H, y\in C;$

(3) $\|(x-y)-(P_{C}x-P_{C}y)\|^{2}\geq\|x-y\|^{2}-\|P_{C}x-P_{C}y\|^{2}, \ \ \ \forall x, y\in H.$

引理 2.8[20] 设 $H$ 是实 Hilbert 空间, $C$ 是 $H$ 上的非空闭凸子集, 令 $T:C\mapsto C$ 是渐近非扩张映射, 如果 $x_{n}\rightharpoonup x$ 且 $x_{n}-Tx_{n}\rightarrow 0$, 那么有 $Tx=x.$

引理 2.9[21] 假设 $\{a_{n}\}$ 是一非负实序列, 使得 $a_{n+1}\leq (1-\delta_{n})a_{n}+\varepsilon_{n},$ 其中 $\{\delta_{n}\}$ 是 (0, 1) 中的序列且 $\{\varepsilon_{n}\}$ 是 $\mathbb{R}$ 中的序列, 满足

(i) $\sum\limits_{n=0}^{\infty}\delta_{n}=\infty;$

(ii) $\limsup\limits_{n\rightarrow\infty}\frac{\varepsilon_{n}}{\delta_{n}}\leq0$ 或者 $\sum\limits_{n=0}^{\infty}|\varepsilon_{n}|<\infty.$

那么 $\lim\limits_{n\rightarrow\infty}a_{n}=0.$

3 主要定理及其证明

定理 3.1 设 $H_{1}$ 和 $H_{2}$ 是两个实 Hilbert 空间, $C_{1}\subset H_{1}$ 和 $C_{2}\subset H_{2}$ 是非空闭凸集. 令$A:H_{1}\mapsto H_{2}$ 是一个有界线性算子, $A^{*}$ 是 $A$ 的伴随算子. 设 $B_{1}, B_{2},\cdots, B_{N}:C_{1}\mapsto H_{1}$ 分别是 $b_{1}, b_{2},\cdots, b_{N}$-逆强单调算子. 假定 $F_{1}, \varphi_{1}:C_{1}\times C_{1}\mapsto \mathbb{R}$ 和 $F_{2}, \varphi_{2}:C_{2}\times C_{2}\mapsto \mathbb{R}$ 是二元函数且满足条件 (A1)-(A7), 其中 $\varphi_{1}, \varphi_{2}$ 是单调的, $\varphi_{1}$ 是弱上半连续的, $F_{2}$ 和 $\varphi_{2}$ 是上半连续的. 设 $S:C_{1}\mapsto C_{1}$ 是一个具有参数 $\theta_{n}$ 的渐近非扩张映射, $f:C_{1}\mapsto C_{1}$ 是一个严格压缩映射, 压缩系数 $0<\rho<1$. 假设 $\Omega= SGEP(F_{1}, \varphi_{1}, F_{2}, \varphi_{2})\cap$ Fix$(G)\cap$ Fix$(S)\neq\emptyset,$ 这里 $G$ 的定义见引理 2.3. 令 $x_{0}\in C_{1}$, 定义序列 $\{x_{n}\}$ 由下式生成

$\begin{cases} v_{n}=T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}, \\ u_{n}=P_{C_{1}}(I-\lambda_{N}B_{N})(av_{n}+(1-a)P_{C_{1}}(I-\lambda_{N-1}B_{N-1}) \\ \ \ \ \ \ \cdots(av_{n}+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots v_{n},\\ x_{n+1}=\alpha_{n}f(x_{n})+\beta_{n}x_{n}+\gamma_{n}S^{n}u_{n}, \ \ \ n\geq0.\end{cases}$

这里 $\{r_{n}\}\subset(r, \infty), r>0, \lambda_{i}\in (0, 2b_{i}), i=1, 2,\cdots, N, \gamma\in (0, \frac{1}{L_{A}}), L_{A}$ 是算子 $A^{*}A$ 的谱半径. $\{\alpha_{n}\}, \{\beta_{n}\}, \{\gamma_{n}\}\subset(0, 1)$ 且$\alpha_{n}+\beta_{n}+\gamma_{n}=1,$ 满足条件

(i) $\lim\limits_{n\rightarrow\infty}\alpha_{n}=0, \sum\limits_{n=0}^{\infty}\alpha_{n}=\infty$;

(ii) $\theta_{n}-1=\eta\alpha_{n}, 0<\eta<1-\rho, \lim\limits_{n\rightarrow\infty}\theta_{n}=1$;

(iii) $\lim\limits_{n\rightarrow\infty}|r_{n+1}-r_{n}|=0, \liminf\limits_{n\rightarrow\infty}r_{n}>0;$

(iv) $0<\liminf\limits_{n\rightarrow\infty}\beta_{n}\leq\limsup\limits_{n\rightarrow\infty}\beta_{n}<1;$

(v) $S$ 满足渐进正则条件, 即 $\lim\limits_{n\rightarrow\infty}\|S^{n+1}x-S^{n}x\|=0$, 对于所有的 $x\in C_{1}$. 那么序列 $\{x_{n}\}$ 强收敛到 $ u_{N}=P_{\Omega}f(u_{N})$.

我们将证明过程分以下五步.

第一步 证明序列 $\{x_{n}\}$ 是有界的. 设 $u_{N}\in\Omega$, 有 $u_{N}=T_{r_{n}}^{(F_{1}, \varphi_{1})}u_{N}$ 和 $Au_{N}=T_{r_{n}}^{(F_{2}, \varphi_{2})}Au_{N}.$

$\|v_{n}-u_{N}\|^{2}=\|T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-T_{r_{n}}^{(F_{1}, \varphi_{1})}u_{N}\|^{2} \notag\\ \leq\|(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-u_{N}\|^{2} \notag\\ \leq\|x_{n}-u_{N}\|^{2}+\gamma^{2}\|A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})Ax_{n}\|^{2} \!+\!2\gamma\langle u_{N}-x_{n}, A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})Ax_{n}\rangle \\ \leq\|x_{n}-u_{N}\|^{2}+\gamma^{2}\langle Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}, AA^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})Ax_{n}\rangle \notag\\ \ +2\gamma\langle Au_{N}-Ax_{n}, (I-T_{r_{n}}^{(F_{2}, \varphi_{2})})Ax_{n}\rangle \notag\\ \leq\|x_{n}-u_{N}\|^{2}+\gamma^{2}L_{A}\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2}+2\gamma\langle Au_{N}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}, Ax_{n} \notag\\ \ -T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\rangle+2\gamma\langle T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}-Ax_{n}, Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\rangle \notag\\ \leq\|x_{n}-u_{N}\|^{2}+\gamma^{2}L_{A}\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2}-2\gamma\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2} \notag\\ \ +2\gamma(\frac{1}{2}\|Au_{N}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2}+\frac{1}{2}\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2}\!-\!\frac{1}{2}\|Au_{N}-Ax_{n}\|^{2}) \notag\\ \leq\|x_{n}-u_{N}\|^{2}+\gamma(\gamma L_{A}-1)\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2},$

因为 $\gamma\in (0, \frac{1}{L_{A}})$, 可知

$\|v_{n}-u_{N}\|\leq\|x_{n}-u_{N}\|,$

由引理 2.4 得到

$\|u_{n}-u_{N}\|\leq\|Gv_{n}-Gu_{N}\|\leq\|v_{n}-u_{N}\|\leq\|x_{n}-u_{N}\|.$

根据 (3.1) 式和条件 (ii) 有

$\|x_{n+1}-u_{N}\|=\|\alpha_{n}(f(x_{n})-u_{N})+\beta_{n}(x_{n}-u_{N})+\gamma_{n}(S^{n}u_{n}-u_{N})\| \notag\\ \leq \alpha_{n}\|f(x_{n})-u_{N}\|+\beta_{n}\|x_{n}-u_{N}\|+\gamma_{n}\theta_{n}\|u_{n}-u_{N}\| \notag\\ \leq \alpha_{n}\rho\|x_{n}-u_{N}\|+\alpha_{n}\|f(u_{N})-u_{N}\|+\beta_{n}\|x_{n}-u_{N}\|+\gamma_{n}\theta_{n}\|x_{n}-u_{N}\| \\ =(\alpha_{n}\rho+\beta_{n}+\gamma_{n}\theta_{n})\|x_{n}-u_{N}\|+\alpha_{n}\|f(u_{N})-u_{N}\| \\ \leq [1-\alpha_{n}(1-\rho-\eta)]\|x_{n}-u_{N}\|+\alpha_{n}(1-\rho-\eta)\frac{\|f(u_{N})-u_{N}\|}{1-\rho-\eta} \\ \leq\max\{\|x_{n}-u_{N}\|, \frac{\|f(u_{N})-u_{N}\|}{1-\rho-\eta}\},$

由数学归纳法, 我们得到

$\|x_{n+1}-u_{N}\|\leq\max\{\|x_{0}-u_{N}\|, \frac{\|f(u_{N})-u_{N}\|}{1-\rho-\eta}\}.$

因此 $\{x_{n}\}$ 是有界的, $\{u_{n}\}, \{v_{n}\}, \{f(x_{n})\}$ 和 $\{S^{n}x_{n}\}$ 也是有界的.

第二步 证明 $\lim\limits_{n\rightarrow\infty}\|x_{n+1}-x_{n}\|=0$ 和 $\lim\limits_{n\rightarrow\infty}\|v_{n}-x_{n}\|=0.$ 由引理 2.6 可知

$\|v_{n+1}-v_{n}\|=\|T_{r_{n+1}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I\!-\!T_{r_{n+1}}^{(F_{2}, \varphi_{2})})A)x_{n+1}\!-\!T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}\| \\ \leq \|T_{r_{n+1}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I\!-\!T_{r_{n+1}}^{(F_{2}, \varphi_{2})})A)x_{n+1}\!-\!T_{r_{n+1}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n+1}}^{(F_{2}, \varphi_{2})})A)x_{n}\| \\ +\|T_{r_{n+1}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n+1}}^{(F_{2}, \varphi_{2})})A)x_{n}\!-\!T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I\!-\!T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}\| \\ \leq\|x_{n+1}-x_{n}\|+\gamma\|A\|\|T_{r_{n+1}}^{(F_{2}, \varphi_{2})}Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|+M_{n} \\ \leq \|x_{n+1}-x_{n}\|+\gamma\|A\|\frac{|r_{n+1}-r_{n}|}{r_{n+1}}\|T_{r_{n+1}}^{(F_{2}, \varphi_{2})}Ax_{n}-Ax_{n}\|+M_{n} \\ =\|x_{n+1}-x_{n}\|+\gamma\|A\|\eta_{n}+M_{n},$

这里

$\eta_{n}=\frac{|r_{n+1}-r_{n}|}{r_{n+1}}\|T_{r_{n+1}}^{(F_{2}, \varphi_{2})}Ax_{n}-Ax_{n}\|;$
$M_{n}=\frac{|r_{n+1}-r_{n}|}{r_{n+1}}\|T_{r_{n+1}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n+1}}^{(F_{2}, \varphi_{2})})A)x_{n}-(I-\gamma A^{*}(I-T_{r_{n+1}}^{(F_{2}, \varphi_{2})})A)x_{n}\|.$

根据序列 $\{u_{n}\}$ 的定义, 我们得到

$\|u_{n+1}-u_{n}\|= \|Gv_{n+1}-Gv_{n}\|\leq\|v_{n+1}-v_{n}\| \leq \|x_{n+1}-x_{n}\|+\gamma\|A\|\eta_{n}+M_{n}.$

设 $x_{n+1}=\beta_{n}x_{n}+(1-\beta_{n})w_{n}$, 结合 (3.1) 式, 我们有

$\|w_{n+1}-w_{n}\|=\ \|\frac{\alpha_{n+1}f(x_{n+1})+\gamma_{n+1}S^{n+1}u_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}f(x_{n})+\gamma_{n}S^{n}u_{n}}{1-\beta_{n}}\| \notag\\=\ \|\frac{\alpha_{n+1}}{1-\beta_{n+1}}(f(x_{n+1})-f(x_{n}))+(\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}})f(x_{n}) \notag\\ +\frac{\gamma_{n+1}}{1-\beta_{n+1}}(S^{n+1}u_{n+1}-S^{n}u_{n})+(\frac{\gamma_{n+1}}{1-\beta_{n+1}}-\frac{\gamma_{n}}{1-\beta_{n}})S^{n}u_{n}\| \notag\\ \leq\ \frac{\rho\alpha_{n+1}}{1-\beta_{n+1}}\|x_{n+1}-x_{n}\|+|\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}}|(\|f(x_{n})\|+\|S^{n}u_{n}\|) \notag\\ +\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n+1}-S^{n+1}u_{n}\|+\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n}-S^{n}u_{n}\| \notag\\ \leq\ \frac{\rho\alpha_{n+1}}{1-\beta_{n+1}}\|x_{n+1}-x_{n}\|+|\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}}|(\|f(x_{n})\|+\|S^{n}u_{n}\|) \notag\\ +\frac{\gamma_{n+1}\theta_{n+1}}{1-\beta_{n+1}}\|u_{n+1}-u_{n}\|+\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n}-S^{n}u_{n}\|,$

将 (3.3) 式代入 (3.4) 式, 我们得到

$\|w_{n+1}-w_{n}\|\leq\ \frac{\rho\alpha_{n+1}+\gamma_{n+1}\theta_{n+1}}{1-\beta_{n+1}}\|x_{n+1}-x_{n}\|+\frac{\gamma_{n+1}\theta_{n+1}}{1-\beta_{n+1}}(\gamma\|A\|\eta_{n}+M_{n}) \notag\\ +|\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}}|(\|f(x_{n})\|+\|S^{n}u_{n}\|) +\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n}-S^{n}u_{n}\| \notag\\ =\ [1-\frac{\alpha_{n+1}(1-\rho-\eta\gamma_{n+1})}{1-\beta_{n+1}}]\|x_{n+1}-x_{n}\|+\frac{\gamma_{n+1}\theta_{n+1}}{1-\beta_{n+1}}(\gamma\|A\|\eta_{n}+M_{n}) \notag\\ +|\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}}|(\|f(x_{n})\|+\|S^{n}u_{n}\|) +\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n}-S^{n}u_{n}\| \notag\\ \leq\ \|x_{n+1}-x_{n}\|+|\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}}|(\|f(x_{n})\|+\|S^{n}u_{n}\|) \notag\\ +\frac{\gamma_{n+1}\theta_{n+1}}{1-\beta_{n+1}}(\gamma\|A\|\eta_{n}+M_{n})+\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n}-S^{n}u_{n}\|,$

从而有

$\|w_{n+1}-w_{n}\|-\|x_{n+1}-x_{n}\|\leq\ |\frac{\alpha_{n+1}}{1-\beta_{n+1}}-\frac{\alpha_{n}}{1-\beta_{n}}|(\|f(x_{n})\|+\|S^{n}u_{n}\|) \notag\\ +\frac{\gamma_{n+1}\theta_{n+1}}{1-\beta_{n+1}}(\gamma\|A\|\eta_{n}+M_{n})+\frac{\gamma_{n+1}}{1-\beta_{n+1}}\|S^{n+1}u_{n}-S^{n}u_{n}\|,$

由条件 (i), (iii), (iv) 和 (v), 我们有

$\limsup\limits_{n\rightarrow\infty}(\|w_{n+1}-w_{n}\|-\|x_{n+1}-x_{n}\|)\leq0.$

根据引理 2.2, 可知

$\lim\limits_{n\rightarrow\infty}\|w_{n}-x_{n}\|=0.$

由序列 $\{w_{n}\}$ 的定义, 我们得到

$\lim\limits_{n\rightarrow\infty}\|x_{n+1}-x_{n}\|=0.$

结合 (3.1) 式, (3.2) 式和条件 (ii), 得到

$\|x_{n+1}-u_{N}\|^{2}\leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\|S^{n}u_{n}-u_{N}\|^{2} \notag\\ \leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\theta_{n}^{2}\|u_{n}-u_{N}\|^{2} \notag\\ \leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\theta_{n}^{2}\|v_{n}-u_{N}\|^{2} \notag\\ \leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\theta_{n}^{2}[\|x_{n}-u_{N}\|^{2} \notag\\ +\gamma(\gamma L_{A}-1)\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2}] \notag\\=\ (\beta_{n}+\gamma_{n}\theta_{n}^{2})\|x_{n}-u_{N}\|^{2}+\alpha_{n}\|f(x_{n})-u_{N}\|^{2} \notag\\ +\gamma_{n}\theta_{n}^{2}\gamma(\gamma L_{A}-1)\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2} \notag\\=\ \|x_{n}-u_{N}\|^{2}+\alpha_{n}\|f(x_{n})-u_{N}\|^{2} +\gamma_{n}\theta_{n}^{2}\gamma(\gamma L_{A}-1)\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2} \notag\\ +\alpha_{n}(\eta^{2}\gamma_{n}\alpha_{n}+2\eta\gamma_{n}-1)\|x_{n}-u_{N}\|^{2},$

移项整理后, 有

$\gamma_{n}\theta_{n}^{2}\gamma(1-\gamma L_{A})\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|^{2} \notag\\ \leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\alpha_{n}(\eta^{2}\gamma_{n}\alpha_{n}+2\eta\gamma_{n}-1)\|x_{n}-u_{N}\|^{2} \notag\\ +\|x_{n+1}-x_{n}\|(\|x_{n+1}-u_{N}\|+\|x_{n}-u_{N}\|),$

由于 $\alpha_{n}\rightarrow0$ 和 $\limsup\limits_{n\rightarrow\infty}\beta_{n}<1,$ 可知 $\liminf\limits_{n\rightarrow\infty}\gamma_{n}>0.$ 再结合 $\theta_{n}\rightarrow1$ 和 (3.5) 式, 我们得到

$\lim\limits_{n\rightarrow\infty}\|Ax_{n}-T_{r_{n}}^{(F_{2}, \varphi_{2})}Ax_{n}\|=0.$

对于 $u_{N}\in\Omega, u_{N}=T_{r_{n}}^{(F_{1}, \varphi_{1})}u_{N}, T_{r_{n}}^{(F_{1}, \varphi_{1})}$ 是定非扩张的, $I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A$ 是非扩张的, 我们有

$\|v_{n}-u_{N}\|^{2}=\ \|T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-T_{r_{n}}^{(F_{1}, \varphi_{1})}u_{N}\|^{2} \\ \leq\ \langle T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-T_{r_{n}}^{(F_{1}, \varphi_{1})}u_{N}, \\ (I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-u_{N}\rangle \\ =\ \langle v_{n}-u_{N}, (I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-u_{N}\rangle \\ =\ \frac{1}{2}\|v_{n}-u_{N}\|^{2}+\frac{1}{2}\|(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n}-u_{N}\|^{2} \\ -\frac{1}{2}\|v_{n}-x_{n}-\gamma A^{*}(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\|^{2} \\ \leq\ \frac{1}{2}\|v_{n}-u_{N}\|^{2}+\frac{1}{2}\|x_{n}-u_{N}\|^{2}-\frac{1}{2}\|v_{n}-x_{n}\|^{2} \\ -\frac{1}{2}\gamma^{2}\|A^{*}(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\|^{2}+\gamma\langle Av_{n}-Ax_{n}, (T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\rangle,$

从而有

$\|v_{n}-u_{N}\|^{2}\leq\ \|x_{n}-u_{N}\|^{2}-\|v_{n}-x_{n}\|^{2}-\gamma^{2}\|A^{*}(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\|^{2} \notag\\ +2\gamma\|Av_{n}-Ax_{n}\|\|(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\| \notag\\ \leq\ \|x_{n}-u_{N}\|^{2}-\|v_{n}-x_{n}\|^{2}+2\gamma\|Av_{n}-Ax_{n}\|\|(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\|,$

将 (3.8) 式代入 (3.6) 式, 得到

$\|x_{n+1}-u_{N}\|^{2}\leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\theta_{n}^{2}(\|x_{n}-u_{N}\|^{2} \\ -\|v_{n}-x_{n}\|^{2}+2\gamma\|Av_{n}-Ax_{n}\|\|(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\|) \\=\ (\beta_{n}+\gamma_{n}\theta_{n}^{2})\|x_{n}-u_{N}\|^{2}+\alpha_{n}\|f(x_{n})-u_{N}\|^{2}-\gamma_{n}\theta_{n}^{2}\|v_{n}-x_{n}\|^{2} \\ +2\gamma\gamma_{n}\theta_{n}^{2}\|Av_{n}-Ax_{n}\|\|(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\| \\=\ \|x_{n}-u_{N}\|^{2}+\alpha_{n}\|f(x_{n})-u_{N}\|^{2}-\gamma_{n}\theta_{n}^{2}\|v_{n}-x_{n}\|^{2} \\ +2\gamma\gamma_{n}\theta_{n}^{2}\|Av_{n}-Ax_{n}\|\|(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\| \\ +\alpha_{n}(\eta^{2}\gamma_{n}\alpha_{n}+2\eta\gamma_{n}-1)\|x_{n}-u_{N}\|^{2},$

移项整理后, 得到

$\gamma_{n}\theta_{n}^{2}\|v_{n}-x_{n}\|^{2}\leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+2\gamma\gamma_{n}\theta_{n}^{2}\|Av_{n}-Ax_{n}\|\|(T_{r_{n}}^{(F_{2}, \varphi_{2})}-I)Ax_{n}\| \\ +\alpha_{n}(\eta^{2}\gamma_{n}\alpha_{n}+2\eta\gamma_{n}-1)\|x_{n}-u_{N}\|^{2} \\ +\|x_{n+1}-x_{n}\|(\|x_{n+1}-u_{N}\|+\|x_{n}-u_{N}\|),$

结合 $\liminf\limits_{n\rightarrow\infty}\gamma_{n}>0, \lim\limits_{n\rightarrow\infty}\alpha_{n}=0, \lim\limits_{n\rightarrow\infty}\theta_{n}=1,$ (3.5) 式和 (3.7) 式, 我们得到

$\lim\limits_{n\rightarrow\infty}\|v_{n}-x_{n}\|=0.$

第三步 证明

$\lim\limits_{n\rightarrow\infty}\|x_{n}-Sx_{n}\|=0.$

$\Phi^{i}x=P_{C_{1}}(I-\lambda_{i}B_{i})(ax+(1-a)P_{C_{1}}(I-\lambda_{i-1}B_{i-1}) \cdots(ax+(1-a)P_{C_{1}}(I-\lambda_{1}B_{1})))\cdots x,$

这里取 $i=\{1, 2,\cdots, N\}$ 和 $\Phi^{0}=I$, 其中 $I$ 是 $H_{1}$ 上的恒等映射. 假设 $u_{N}\in\Omega,$ 那么 $\Phi^{N}u_{N}=u_{N}.$ 运用引理 2.1, 我们有

$\|u_{n}-u_{N}\|^{2}=\ \|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2}\\ =\ \|P_{C_{1}}(I-\lambda_{N}B_{N})(aI+(1-a)\Phi^{N-1})v_{n} \\ -P_{C_{1}}(I-\lambda_{N}B_{N})(aI+(1-a)\Phi^{N-1})u_{N}\|^{2} \\ \leq\ \|(aI+(1-a)\Phi^{N-1})v_{n}-(aI+(1-a)\Phi^{N-1})u_{N}\|^{2}+\lambda_{N}(\lambda_{N}-2b_{N}) \\ \cdot\|B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n})-B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N})\|^{2} \\ \leq\ a\|v_{n}-u_{N}\|^{2}+(1-a)\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}\|^{2}+\lambda_{N}(\lambda_{N}-2b_{N})\|B_{N}(av_{n} \\ +(1-a)\Phi^{N-1}v_{n})-B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N})\|^{2} \\ \leq\ a\|v_{n}-u_{N}\|^{2}+(1-a)[a\|v_{n}-u_{N}\|^{2}+(1-a)\|\Phi^{N-2}v_{n}-\Phi^{N-2}u_{N}\|^{2} \\ +\lambda_{N-1}(\lambda_{N-1}-2b_{N-1})\|B_{N-1}(av_{n}+(1-a)\Phi^{N-2}v_{n})-B_{N-1}(au_{N} \\ +(1-a)\Phi^{N-2}u_{N})\|^{2}]+\lambda_{N}(\lambda_{N}-2b_{N})\|B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n}) \\ -B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N})\|^{2},$

因此, 由数学归纳法可以得到

$\|u_{n}-u_{N}\|^{2}\leq\ \|v_{n}-u_{N}\|^{2}+\sum\limits_{i=1}^{N}\lambda_{i}(\lambda_{i}-2b_{i})\|B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n}) \notag\\ -B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})\|^{2} \notag\\ \leq\ \|x_{n}-u_{N}\|^{2}+\sum\limits_{i=1}^{N}\lambda_{i}(\lambda_{i}-2b_{i})\|B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})\notag\\ -B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})\|^{2},$

结合 (3.6) 式和 (3.10) 式, 我们有

$\|x_{n+1}-u_{N}\|^{2}\leq\ \alpha_{n}\|f(x_{n}-u_{N})\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\theta_{n}^{2}[\|x_{n}-u_{N}\|^{2}\\ +\sum\limits_{i=1}^{N}\lambda_{i}(\lambda_{i}-2b_{i}) \|B_{i}(av_{n}\!+\!(1-a)\Phi^{i-1}v_{n})\!-\!B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})\|^{2}],$

移项整理后, 可知

$-\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\lambda_{i}(\lambda_{i}-2b_{i})\|B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})-B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})\|^{2}\\ \leq\ \alpha_{n}\|f(x_{n}-u_{N})\|^{2}+\|x_{n+1}-x_{n}\|(\|x_{n}-u_{N}\|+\|x_{n+1}-u_{N}\|)\\ +\alpha_{n}(\eta^{2}\gamma_{n}\alpha_{n}+2\eta\gamma_{n}-1)\|x_{n}-u_{N}\|^{2}, $

结合 $\liminf\limits_{n\rightarrow\infty}\gamma_{n}>0, \lim\limits_{n\rightarrow\infty}\alpha_{n}=0$ 和 (3.5) 式, 我们有

$\lim\limits_{n\rightarrow\infty}\|B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})-B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})\|=0.$

利用引理 2.7, 得到

$\|u_{n}-u_{N}\|^{2}=\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2}\\ =\ \|P_{C_{1}}(I-\lambda_{N}B_{N})(aI+(1-a)\Phi^{N-1})v_{n} -P_{C_{1}}(I-\lambda_{N}B_{N})(aI+(1-a)\Phi^{N-1})u_{N}\|^{2}\\ \leq\ a\langle v_{n}-u_{N}, \Phi^{N}v_{n}-\Phi^{N}u_{N}\rangle+(1-a)\langle\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}, \Phi^{N}v_{n}-\Phi^{N}u_{N}\rangle\\ +\lambda_{N}\langle B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N}) -B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n}), \Phi^{N}v_{n}-\Phi^{N}u_{N}\rangle\\ \leq\ \frac{a}{2}[\|v_{n}-u_{N}\|^{2}+\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2}-\|v_{n}-u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2}] +\frac{1-a}{2}[\|\Phi^{N-1}v_{n} \\ -\Phi^{N-1}u_{N}\|^{2}+\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2}-\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2}]\\ +\lambda_{N}\langle B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N})-B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n}), \Phi^{N}v_{n}-\Phi^{N}u_{N}\rangle \\ =\ \frac{a}{2}\|v_{n}-u_{N}\|^{2}+\frac{1-a}{2}\|\Phi^{N-1}v_{n}+\Phi^{N-1}u_{N}\|^{2}+\frac{1}{2}\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2} \\ -\frac{1-a}{2}\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2} \\ +\lambda_{N}\langle B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N})-B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n}), \Phi^{N}v_{n}-\Phi^{N}u_{N}\rangle\\ -\frac{a}{2}\|v_{n}-u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2},$

移项整理后, 得到

$\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2} \leq\ a\|v_{n}-u_{N}\|^{2}+(1-a)\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}\|^{2} \\ -a\|v_{n}-u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2} +2\lambda_{N}\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\| \\ \cdot \|B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N})-B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n})\| \notag\\ -(1-a)\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2},$

类似可得

$\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}\|^{2} \leq\ a\|v_{n}-u_{N}\|^{2}+(1-a)\|\Phi^{N-2}v_{n}-\Phi^{N-2}u_{N}\|^{2}\ \\ -a\|v_{n}-u_{N}-\Phi^{N-1}v_{n}+\Phi^{N-1}u_{N}\|^{2}\notag\\ -(1-a)\|\Phi^{N-2}v_{n}-\Phi^{N-2}u_{N}-\Phi^{N-1}v_{n}+\Phi^{N-1}u_{N}\|^{2} \\ +2\lambda_{N-1}\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}\| \|B_{N-1}(au_{N}+(1-a)\Phi^{N-2}u_{N}) \\ -B_{N-1}(av_{n}+(1-a)\Phi^{N-2}v_{n})\|,$

结合 (3.12) 式和 (3.13) 式, 有

$\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2} \leq\ a\|v_{n}-u_{N}\|^{2}+(1-a)[a\|v_{n}-u_{N}\|^{2} -a\|v_{n}-u_{N}\\ -\Phi^{N-1}v_{n}+\Phi^{N-1}u_{N}\|^{2} +(1-a)\|\Phi^{N-2}v_{n}-\Phi^{N-2}u_{N}\|^{2}\\ -(1-a)\|\Phi^{N-2}v_{n}-\Phi^{N-2}u_{N}-\Phi^{N-1}v_{n}+\Phi^{N-1}u_{N}\|^{2} \\ +2\lambda_{N-1}\|B_{N-1}(au_{N}+(1-a)\Phi^{N-2}u_{N})\\ -B_{N-1}(av_{n}+(1-a)\Phi^{N-2}v_{n})\|\|\Phi^{N-1}v_{n} -\Phi^{N-1}u_{N}\|]\\ -a\|v_{n}-u_{N}-\Phi^{N}v_{n}+\Phi^{N}u_{N}\|^{2}-(1-a)\|\Phi^{N-1}v_{n}-\Phi^{N-1}u_{N}\\ -\Phi^{N}v_{n} +\Phi^{N}u_{N}\|^{2}+2\lambda_{N}\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|\\ \cdot \|B_{N}(au_{N}+(1-a)\Phi^{N-1}u_{N}) -B_{N}(av_{n}+(1-a)\Phi^{N-1}v_{n})\|,$

由数学归纳法可知

$\|\Phi^{N}v_{n}-\Phi^{N}u_{N}\|^{2}\leq\ \|v_{n}-u_{N}\|^{2}-a\sum\limits_{i=1}^{N}\|v_{n}-u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2}+2\sum\limits_{i=1}^{N}\lambda_{i}\|\Phi^{i}v_{n} \notag\\ -\Phi^{i}u_{N}\|\|B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})-B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})\| \notag\\ -(1-a)\sum\limits_{i=1}^{N}\|\Phi^{i-1}v_{n}-\Phi^{i-1}u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2},$

将 (3.14) 式代入 (3.6) 式, 得到

$\|x_{n+1}-u_{N}\|^{2}\leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\beta_{n}\|x_{n}-u_{N}\|^{2}+\gamma_{n}\theta_{n}^{2}[\|v_{n}-u_{N}\|^{2} \\ -(1-a)\sum\limits_{i=1}^{N}\|\Phi^{i-1}v_{n}-\Phi^{i-1}u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2} \\ +2\sum\limits_{i=1}^{N}\lambda_{i}\|\Phi^{i}v_{n}-\Phi^{i}u_{N}\|\|B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N}) \\ -B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})\| -a\sum\limits_{i=1}^{N}\|v_{n}-u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2}] \\ \leq\ (\beta_{n}+\gamma_{n}\theta_{n}^{2})\|x_{n}-u_{N}\|^{2}+\alpha_{n}\|f(x_{n})-u_{N}\|^{2}\\ -a\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\|v_{n}-u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2}\\ +2\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\lambda_{i}\|B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})\\ -B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})\|\|\Phi^{i}v_{n}-\Phi^{i}u_{N}\|\\ -(1-a)\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\|\Phi^{i-1}v_{n}-\Phi^{i-1}u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2},$

这表明

$a\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\|v_{n}-u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2}\\ +(1-a)\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\|\Phi^{i-1}v_{n}-\Phi^{i-1}u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|^{2}\\ \leq\ \alpha_{n}\|f(x_{n})-u_{N}\|^{2}+\|x_{n+1}-x_{n}\|(\|x_{n+1}-u_{N}\|+\|x_{n}-u_{N}\|)\\ +\alpha_{n}(\eta^{2}\gamma_{n}\alpha_{n}+2\eta\gamma_{n}-1)\|x_{n}-u_{N}\|^{2}+2\gamma_{n}\theta_{n}^{2}\sum\limits_{i=1}^{N}\lambda_{i}\|\Phi^{i}v_{n}-\Phi^{i}u_{N}\|\\ \cdot\|B_{i}(au_{N}+(1-a)\Phi^{i-1}u_{N})-B_{i}(av_{n}+(1-a)\Phi^{i-1}v_{n})\|,$

由于 $\liminf\limits_{n\rightarrow\infty}\gamma_{n}>0, \lim\limits_{n\rightarrow\infty}\alpha_{n}=0, \lim\limits_{n\rightarrow\infty}\theta_{n}=1$, 再结合 (3.5) 式和 (3.11) 式, 我们有

$\lim\limits_{n\rightarrow\infty}\|v_{n}-u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|=0; \notag\\ \lim\limits_{n\rightarrow\infty}\|\Phi^{i-1}v_{n}-\Phi^{i-1}u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|=0.$

根据 (3.15) 式, 可得

$\|v_{n}-u_{n}\|=\|\Phi^{0}v_{n}-\Phi^{N}v_{n}\| \leq\sum\limits_{i=1}^{N}\|\Phi^{i-1}v_{n}-\Phi^{i-1}u_{N}-\Phi^{i}v_{n}+\Phi^{i}u_{N}\|\rightarrow0(n\rightarrow\infty),$

即 $\lim\limits_{n\rightarrow\infty}\|G(v_{n})-v_{n}\|=0.$ 由引理 2.4 可知, $G$ 是非扩张映射, 那么有

$\|G(x_{n})-x_{n}\|\leq\|G(x_{n})-G(v_{n})\|+\|G(v_{n})-v_{n}\|+\|v_{n}-x_{n}\| \notag\\ \leq2\|v_{n}-x_{n}\|+\|G(v_{n})-v_{n}\|\rightarrow0(n\rightarrow\infty),$

由 (3.5) 式, (3.9) 式和 (3.16) 式得

$\|x_{n+1}-u_{n}\|\leq\|x_{n+1}-x_{n}\|+\|x_{n}-v_{n}\|+\|v_{n}-u_{n}\|,$

$\lim\limits_{n\rightarrow\infty}\|x_{n+1}-u_{n}\|=0.$

对于每一个 $n\geq0$, 有

$\|x_{n+1}-S^{n}u_{n}\|=\|\alpha_{n}f(x_{n})+\beta_{n}x_{n}+\gamma_{n}S^{n}u_{n}-S^{n}u_{n}\|\\ \leq\alpha_{n}\|f(x_{n})-S^{n}u_{n}\|+\beta_{n}\|x_{n+1}-x_{n}\|+\beta_{n}\|x_{n+1}-S^{n}u_{n}\|,$

进一步有

$\|x_{n+1}-S^{n}u_{n}\|\leq\frac{\alpha_{n}}{1-\beta_{n}}\|f(x_{n})-S^{n}u_{n}\|+\frac{\beta_{n}}{1-\beta_{n}}\|x_{n+1}-x_{n}\|,$

根据条件 (i), 条件 (iv) 和 (3.5) 式, 得到

$\lim\limits_{n\rightarrow\infty}\|x_{n+1}-S^{n}u_{n}\|=0.$

观察

$\|x_{n}-S^{n}x_{n}\|\leq\|x_{n}-x_{n+1}+x_{n+1}-S^{n}u_{n}+S^{n}u_{n}-S^{n}x_{n}\|\\ \leq\|x_{n}-x_{n+1}\|+\|x_{n+1}-S^{n}u_{n}\|+\theta_{n}(\|u_{n}-x_{n+1}\|+\|x_{n+1}-x_{n}\|),$

由 (3.5) 式, (3.18) 式和 (3.19) 式, 得到

$\lim\limits_{n\rightarrow\infty}\|x_{n}-S^{n}x_{n}\|=0.$

由于 $S$ 是渐近非扩张映射, 可知

$\|x_{n}-Sx_{n}\|\leq\|x_{n}-x_{n+1}\|\!+\!\|x_{n+1}-S^{n+1}x_{n+1}\|\!+\!\|S^{n+1}x_{n+1}-S^{n+1}x_{n}\|\! +\!\|S^{n+1}x_{n}-Sx_{n}\|\\ \leq\|x_{n}-x_{n+1}\|+\|x_{n+1}-S^{n+1}x_{n+1}\|+\theta_{n+1}\|x_{n}-x_{n+1}\|+\theta_{1}\|S^{n}x_{n}-x_{n}\|,$

结合 (3.5) 式和 (3.20) 式, 有

$\lim\limits_{n\rightarrow\infty}\|x_{n}-Sx_{n}\|=0.$

第四步 证明 $\limsup\limits_{n\rightarrow\infty}\langle f(u_{N})-u_{N}, x_{n}-u_{N}\rangle\leq0, u_{N}=P_{\Omega}f(u_{N}).$ 由于$\{x_{n}\}$ 是有界的, 取 $\{x_{n}\}$ 的子序列 $\{x_{n_{j}}\}$, 使得

$\limsup\limits_{n\rightarrow\infty}\langle f(u_{N})-u_{N}, x_{n}-u_{N}\rangle=\limsup\limits_{j\rightarrow\infty}\langle f(u_{N})-u_{N}, x_{n_{j}}-u_{N}\rangle.$

当 $\{x_{n_{j}}\}$ 是有界的, 存在子序列 $\{x_{n_{j_{i}}}\}\subset\{x_{n_{j}}\}$ 弱收敛到某一点 $w\in C_{1}$. 不失一般性, 设 $x_{n_{j}}\rightharpoonup w$.

下证 $w\in\Omega.$ 首先, 证明 $w\in $ Fix$(S)\bigcap $ Fix$(G)$, 结合 (3.17) 式, (3.21) 式和 $x_{n_{j}}\rightharpoonup w,$ 利用引理 2.8 可知 $w\in $ Fix$(S)\bigcap $ Fix$(G)$. 接着证明 $w\in SGEP(F_{1}, \varphi_{1}, F_{2}, \varphi_{2}),$ 根据 (3.1) 式定义, 有

$v_{n}=T_{r_{n}}^{(F_{1}, \varphi_{1})}(I-\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})A)x_{n},$

$F_{1}(v_{n}, y)+\varphi_{1}(v_{n}, y)+\frac{1}{r_{n}}\langle y-v_{n}, v_{n}-x_{n}+\gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})Ax_{n}\rangle\geq0, \ \ \ \forall y\in C_{1}.$

由于 $F_{1}$ 和 $\varphi_{1}$ 是单调的, 可得

$\frac{1}{r_{n}}\langle y-v_{n}, v_{n}-x_{n}\rangle+\frac{1}{r_{n}}\langle y-v_{n}, \gamma A^{*}(I-T_{r_{n}}^{(F_{2}, \varphi_{2})})Ax_{n}\rangle\geq F_{1}(y, v_{n})+\varphi_{1}(y, v_{n}), \ \ \ \forall y\in C_{1}.$

用 $n_{j}$ 代替上式的 $n$, 我们有

$\frac{1}{r_{n_{j}}}\langle y-v_{n_{j}}, v_{n_{j}}-x_{n_{j}}\rangle+\frac{1}{r_{n_{j}}}\langle y-v_{n_{j}}, \gamma A^{*}(I-T_{r_{n_{j}}}^{(F_{2}, \varphi_{2})})Ax_{n_{j}}\rangle\geq F_{1}(y, v_{n_{j}})+\varphi_{1}(y, v_{n_{j}}),$

由于 $\|v_{n_{j}}-x_{n_{j}}\|\rightarrow0, \|A^{*}(I-T_{r_{n_{j}}}^{(F_{2}, \varphi_{2})})Ax_{n_{j}}\|\rightarrow0$ 和 $\|x_{n_{j}}-w\|\rightarrow0, j\rightarrow\infty.$ 并结合条件 (iii), 条件 (A4) 和条件 (A7), 我们得到

$F_{1}(y, w)+\varphi_{1}(y, w)\leq0, \ \ \ \forall y\in C_{1}.$

设 $y_{t}=ty+(1-t)w, \forall t\in(0, 1], \forall y\in C_{1}$. 可知 $y_{t}\in C_{1}$, 且有

$F_{1}(y_{t}, w)+\varphi_{1}(y_{t}, w)\leq0.$

根据条件 (A1)-(A7), 我们得到

$0\leq F_{1}(y_{t}, y_{t})+\varphi_{1}(y_{t}, y_{t}) \\ \leq t(F_{1}(y_{t}, y)+\varphi_{1}(y_{t}, y))+(1-t)(F_{1}(y_{t}, w)+\varphi_{1}(y_{t}, w)) \\ \leq t(F_{1}(y_{t}, y)+\varphi_{1}(y_{t}, y)) \\ \leq F_{1}(y_{t}, y)+\varphi_{1}(y_{t}, y), \ \ \ \forall y\in C_{1}.$

运用条件 (A3) 和 $\varphi_{1}$ 是弱上半连续的, 当 $t\rightarrow0$ 时, 我们有 $F_{1}(w, y)+\varphi_{1}(w, y)\geq0, \forall y\in C_{1}.$ 即 $w\in GEP(F_{1}, \varphi_{1}).$

下证 $Aw\in GEP(F_{2}, \varphi_{2}).$ 由于 $A$ 是有界线性算子, 那么 $Ax_{n_{j}}\rightarrow Aw.$ 设

$z_{n_{j}}=Ax_{n_{j}}-T_{r_{n_{j}}}^{(F_{2}, \varphi_{2})}Ax_{n_{j}},$

结合 (3.7) 式, 可知 $\lim\limits_{n\rightarrow\infty}z_{n_{j}}=0$ 和 $T_{r_{n_{j}}}^{(F_{2}, \varphi_{2})}Ax_{n_{j}}=Ax_{n_{j}}-z_{n_{j}}.$ 对于 $\forall z\in C_{2}$, 利用引理 2.5, 我们有

$F_{2}(Ax_{n_{j}}-z_{n_{j}}, z)+\varphi_{2}(Ax_{n_{j}}-z_{n_{j}}, z)+\frac{1}{r_{n_{j}}}\langle z-(Ax_{n_{j}}-z_{n_{j}}), (Ax_{n_{j}}-z_{n_{j}})-Ax_{n_{j}}\rangle\geq 0,$

由于 $F_{2}$ 和 $\varphi_{2}$ 是上半连续的, 得到 $F_{2}(Aw, z)+\varphi_{2}(Aw, z)\geq0, \forall z\in C_{2}.$ 即 $Aw\in GEP(F_{2}, \varphi_{2})$, 所以 $w\in SGEP(F_{1}, \varphi_{1}, F_{2}, \varphi_{2}).$ 综上可知, $w\in\Omega.$

已知 $u_{N}=P_{\Omega}f(u_{N})$ 和 $x_{n_{j}}\rightharpoonup w$, 可得

$\limsup\limits_{n\rightarrow\infty}\langle f(u_{N})-u_{N}, x_{n}-u_{N}\rangle=\limsup\limits_{j\rightarrow\infty}\langle f(u_{N})-u_{N}, x_{n_{j}}-u_{N}\rangle \notag\\ =\langle f(u_{N})-u_{N}, w-u_{N}\rangle\leq0.$

第五步 证明序列强收敛到 $u_{N}$. 考虑

$\|x_{n+1}-u_{N}\|^{2}=\langle \alpha_{n}f(x_{n})+\beta_{n}x_{n}+\gamma_{n}S^{n}u_{n}-u_{N}, x_{n+1}-u_{N}\rangle \\ =\alpha_{n}\langle f(x_{n})-f(u_{N}), x_{n+1}-u_{N}\rangle+\alpha_{n}\langle f(u_{N})-u_{N}, x_{n+1}-u_{N}\rangle \\ +\beta_{n}\langle x_{n}-u_{N}, x_{n+1}-u_{N}\rangle+\gamma_{n}\langle S^{n}u_{n}-u_{N}, x_{n+1}-u_{N}\rangle\\ \leq\alpha_{n}\rho\|x_{n}-u_{N}\|\|x_{n+1}-u_{N}\|+\beta_{n}\|x_{n}-u_{N}\|\|x_{n+1}-u_{N}\| \\ +\gamma_{n}\theta_{n}\|x_{n}-u_{N}\|\|x_{n+1}-u_{N}\|+\alpha_{n}\langle f(u_{N})-u_{N}, x_{n+1}-u_{N}\rangle \\=(\alpha_{n}\rho+\beta_{n}+\gamma_{n}\theta_{n})\|x_{n}-u_{N}\|\|x_{n+1}-u_{N}\|+\alpha_{n}\langle f(u_{N})-u_{N}, x_{n+1}-u_{N}\rangle\\ \leq\frac{\alpha_{n}\rho+\beta_{n}+\gamma_{n}\theta_{n}}{2}(\|x_{n}\!-\!u_{N}\|^{2}+\|x_{n+1}\!-\!u_{N}\|^{2})\!+\!\alpha_{n}\langle f(u_{N})\!-\!u_{N}, x_{n+1}\!-\!u_{N}\rangle,$

这表明

$\|x_{n+1}-u_{N}\|^{2}\leq\ [1-\frac{2\alpha_{n}(1-\rho-\eta)}{1+\alpha_{n}(1-\rho-\eta)}]\|x_{n}-u_{N}\|^{2} \\ +\frac{2\alpha_{n}(1-\rho-\eta)}{1+\alpha_{n}(1-\rho-\eta)}\cdot\frac{\langle f(u_{N})-u_{N}, x_{n+1}-u_{N}\rangle}{1-\rho-\eta}.$

$\delta_{n}=\frac{2\alpha_{n}(1-\rho-\eta)}{1+\alpha_{n}(1-\rho-\eta)}, \varepsilon_{n}=\frac{2\alpha_{n}(1-\rho-\eta)}{1+\alpha_{n}(1-\rho-\eta)}\cdot\frac{\langle f(u_{N})-u_{N}, x_{n+1}-u_{N}\rangle}{1-\rho-\eta},$

由条件 (i) 和 (3.22) 式, 有

$\sum\limits_{n=0}^{\infty}\delta_{n}=\sum\limits_{n=0}^{\infty}\frac{2\alpha_{n}(1-\rho-\eta)}{1+\alpha_{n}(1-\rho-\eta)}\leq\sum\limits_{n=0}^{\infty}2\alpha_{n}(1-\rho-\eta)=+\infty,$
$\limsup\limits_{n\rightarrow\infty}\frac{\varepsilon_{n}}{\delta_{n}}=\limsup\limits_{n\rightarrow\infty}\frac{\langle f(u_{N})-u_{N}, x_{n+1}-u_{N}\rangle}{1-\rho-\eta}\leq0.$

应用引理 2.9, 得到 $\lim\limits_{n\rightarrow\infty}\|x_{n+1}-u_{N}\|=0.$

因此, 序列 $\{x_{n}\}$ 强收敛到 $u_{N}=P_{\Omega}f(u_{N})$. 再应用引理 2.3, 得 $(u_{1}, u_{2},\cdots, u_{N})$ 是问题 (1.6) 的一个解, 其中 $u_{i}=P_{C_{1}}(I-\lambda_{i}B_{i})(au_{N}+(1-a)u_{i-1}), u_{1}=P_{C_{1}}(I-\lambda_{1}B_{1})u_{N}, $ $i=2, 3, \cdots, N.$ 定理 3.1 得证.

推论 3.1 设 $H_{1}$ 和 $H_{2}$ 是两个实 Hilbert 空间, $C_{1}\subset H_{1}$ 和 $C_{2}\subset H_{2}$ 是非空闭凸集. 令 $A:H_{1}\mapsto H_{2}$ 是一个有界线性算子, $A^{*}$是 $A$ 的伴随算子. 设 $B_{1}:C_{1}\mapsto H_{1}$ 是 $b_{1}$-逆强单调算子. 假定 $F_{1}:C_{1}\times C_{1}\mapsto \mathbb{R}$ 和 $F_{2}:C_{2}\times C_{2}\mapsto R$ 是二元函数且满足条件 (A1)-(A4), 其中 $F_{2}$ 是上半连续的. 令 $S:C_{1}\mapsto C_{1}$是一个非扩张映射, $f:C_{1}\mapsto C_{1}$ 是一个严格压缩映射, 压缩系数 $0<\rho<1$. 假设 $\Omega= SEP(F_{1}, F_{2})\cap $ Fix$(G)\cap $ Fix$(S)\neq\emptyset,$ 取 $x_{0}\in C_{1}$, 定义序列 $\{x_{n}\}$ 由下式生成

$\begin{cases} v_{n}=T_{r_{n}}^{F_{1}}(I-\gamma A^{*}(I-T_{r_{n}}^{F_{2}})A)x_{n},\\ u_{n}=P_{C_{1}}(I-\lambda_{1}B_{1})v_{n}, \\ x_{n+1}=\alpha_{n}f(x_{n})+\beta_{n}x_{n}+\gamma_{n}Su_{n}, \ \ \ n\geq0. \end{cases}$

这里 $\{r_{n}\}\subset(r, \infty), r>0, \lambda_{1}\in (0, 2b_{1}), \gamma\in (0, \frac{1}{L_{A}}), L_{A}$ 是算子 $A^{*}A$ 的谱半径. $\{\alpha_{n}\}, \{\beta_{n}\},$ $\{\gamma_{n}\}\subset(0, 1)$ 且 $\alpha_{n}+\beta_{n}+\gamma_{n}=1,$ 满足如下条件

(i) $\lim\limits_{n\rightarrow\infty}\alpha_{n}=0, \sum\limits_{n=0}^{\infty}\alpha_{n}=\infty$;

(ii) $\lim\limits_{n\rightarrow\infty}|r_{n+1}-r_{n}|=0, \liminf\limits_{n\rightarrow\infty}r_{n}>0;$

(iii) $0<\liminf\limits_{n\rightarrow\infty}\beta_{n}\leq\limsup\limits_{n\rightarrow\infty}\beta_{n}<1;$

那么序列 $\{x_{n}\}$ 强收敛到 $u=P_{\Omega}f(u)$.

在定理 3.1 中, 令 $\varphi_{1}=\varphi_{2}=0, N=1, a=0, S$ 是非扩张映射, 推论得证.

参考文献

Chang S S, Lee H, Chi K C.

A new method for solving equilibrium problem fixed point problem and variational inequality problem with application to optimization

Nonlinear Analysis, 2009, 70(9): 3307-3319

DOI:10.1016/j.na.2008.04.035      URL     [本文引用: 1]

Plubtieng S, Punpaeng R.

A general iterative method for equilibrium problems and fixed point problems in Hilbert spaces

Journal of Mathematical Analysis and Applications, 2007, 336(1): 455-469

DOI:10.1016/j.jmaa.2007.02.044      URL     [本文引用: 1]

Sitthithakerngkiet K, Deepho J, Martínez-Moreno, et al.

An iterative approximation scheme for solving a split generalized equilibrium, variational inequalities and fixed point problems

International Journal of Computer Mathematics, 2017, 94(12): 2373-2395

DOI:10.1080/00207160.2017.1283409      URL     [本文引用: 1]

Jaiboon C, Chantarangsi W, Kumam P.

A convergence theorem based on a hybrid relaxed extragradient method for generalized equilibrium problems and fixed point problems of a finite family of nonexpansive mappings

Nonlinear Analysis Hybrid Systems, 2010, 4(1): 199-215

DOI:10.1016/j.nahs.2009.09.009      URL     [本文引用: 1]

Takahashi S, Takahashi W.

Viscosity approximation methods for equilibrium problems and fixed point problems in Hilbert spaces

Journal of Mathematical Analysis and Applications, 2007, 331(1): 506-515

DOI:10.1016/j.jmaa.2006.08.036      URL     [本文引用: 1]

Plubtieng S, Thammathiwat T.

A viscosity approximation method for finding a common solution of fixed points and equilibrium problems in Hilbert spaces

Journal of Global Optimization, 2011, 50(2): 313-327

DOI:10.1007/s10898-010-9583-z      URL     [本文引用: 1]

Takahashi S, Takahashi W.

Strong convergence theorem for a generalized equilibrium problem and a nonexpansive mapping in a Hilbert space

Nonlinear Analysis, 2008, 69(3): 1025-1033

DOI:10.1016/j.na.2008.02.042      URL     [本文引用: 1]

Wang S, Gong X, Kang S.

Strong convergence theorem on split equilibrium and fixed point problems in Hilbert spaces

The Bulletin of the Malaysian Mathematical Society Series 2, 2016, 41(3): 1309-1326

[本文引用: 1]

Phuengrattana W, Klanarong C.

Strong convergence of the viscosity approximation method for the split generalized equilibrium problem

Rendiconti del Circolo Matematico di Palermo Series 2, 2021, 71: 39-64

DOI:10.1007/s12215-021-00617-7      [本文引用: 2]

Yao Y, Leng L, Postolache M, et al.

Mann type iteration method for solving the split common fixed point problem

J Nonlinear Convex Anal, 2017, 18(5): 875-882

[本文引用: 1]

Wu X, Zhao L.

Viscosity approximation methods for multivalued nonexpansive mappings

Mediterr J Math, 2016, 13(5): 2645-2657

DOI:10.1007/s00009-015-0644-x      URL     [本文引用: 1]

Yao Y, Postolache M, Yao J C.

An iterative algorithm for solving the generalized variational inequalities and fixed points problems

Mathematics, 2019, 7(1): 61

DOI:10.3390/math7010061      URL     [本文引用: 1]

In this paper, a generalized variational inequality and fixed points problem is presented. An iterative algorithm is introduced for finding a solution of the generalized variational inequalities and fixed point of two quasi-pseudocontractive operators under a nonlinear transformation. Strong convergence of the suggested algorithm is demonstrated.

Cheawchan K, Kangtunyakarn A.

The modified split generalized equilibrium problem for quasi-nonexpansive mappings and applications

Journal of Inequalities and Applications, 2018, 2018: Article number 122

[本文引用: 1]

Zhao Y, Liu X, Sun R.

Iterative algorithms of common solutions for a hierarchical fixed point problem, a system of variational inequalities, and a split equilibrium problem in Hilbert spaces

Journal of Inequalities and Applications, 2021, 2021: Article number 11

[本文引用: 2]

夏平静, 蔡钢.

Hilbert 空间中变分不等式问题的自适应粘性算法

数学物理学报, 2023, 43A(2): 581-592

[本文引用: 1]

Xia P J, Cai G.

Self adaptive viscosity algorithm for solving variational inequality problem in Hilbert spaces

Acta Mathematica Scientia, 2023, 43A(2): 581-592

[本文引用: 1]

Ceng L C, Wang C Y, Yao J C.

Strong convergence theorems by a relaxed extragradient method for a general system of variational inequalities

Journal of Inequalities and Applications, 2008, 67(3): 375-390

[本文引用: 1]

Wang Y H, Pan C J.

Viscosity approximation methods for a general variational inequality system and fixed point problems in Banach spaces

Symmetry, 2020, 12(1): 36

DOI:10.3390/sym12010036      URL     [本文引用: 1]

In Banach spaces, we study the problem of solving a more general variational inequality system for an asymptotically non-expansive mapping. We give a new viscosity approximation scheme to find a common element. Some strong convergence theorems of the proposed iterative method are obtained. A numerical experiment is given to show the implementation and efficiency of our main theorem. Our results presented in this paper generalize and complement many recent ones.

Suzuki, Tomonari.

Strong convergence theorems for infinite families of nonexpansive mappings in general Banach spaces

Fixed Point Theory and Applications, 2005, 1: 103-123

[本文引用: 1]

Filomena C, Giuseppe M, Luigi M, et al.

A hybrid projection algorithm for finding solutions of mixed equilibrium problem and variational inequality problem

Fixed Point Theory and Applications, 2010, 2010: Article number 383740

[本文引用: 1]

Jecho Y, Zhou H, Guo G.

Weak and strong convergence theorems for three-step iterations with errors for asymptotically nonexpansive mappings

Computers and Mathematics with Applications, 2004, 47(4/5): 707-717

DOI:10.1016/S0898-1221(04)90058-2      URL     [本文引用: 1]

Xu H K.

Another control condition in an iterative method for nonexpansive mappings

Bulletin of the Australian Mathematical Society, 2002, 65(1): 109-113

DOI:10.1017/S0004972700020116      URL     [本文引用: 1]

We prove the convergence of an iterative method to a fixed point of a nonexpansive mapping in a uniformly smooth Banach space. We are able to relax one of the control conditions of P.L. Lions (1977).

杨静, 龙宪军.

关于伪单调变分不等式与不动点问题的新投影算法

数学物理学报, 2022, 42A(3): 904-919

[本文引用: 1]

Yang J, Long X J.

A new projection algorithm for solving pseudo-monotone variational inequality and fixed point problems

Acta Mathematica Scientia, 2022, 42A(3): 904-919

[本文引用: 1]

/