数学物理学报  2016, Vol. 36 Issue (4): 763-770   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
邓伟奇
关于二元函数列的分裂均衡问题系统*
邓伟奇     
云南财经大学统计与数学学院 昆明 650221
摘要:旨在引入一全新方法,用以逼近关于非线性算子族的分裂公共不动点问题的解;并借此构造迭代算法,以求解一类新型问题---关于二元函数列的分裂均衡问题系统,且于Hilbert空间背景下建立强收敛定理.
关键词分裂均衡问题系统     公共不动点     Hilbert空间    
A System of Split Equilibrium Problems for Sequences of Bifunctions
Deng Weiqi     
School of Statistics and Mathematics, Yunnan University of Finance and Economics, Kunming 650221
Abstract: The purpose in this paper is to introduce an up-to-date method for the approximation of an element of the set of solutions to the split common fixed point problem for countable families of nonlinear operators, by which an iterative algorithm is developed for solving a new type of problems, namely, a system of split equilibrium problems for two sequences of bifunctions. A strong convergence theorem is established in the framework of Hilbert spaces.
Key words: System of split equilibrium problems     Common fixed points     Hilbert spaces    
1 引言

本文始终假设$H_1$$H_2$为赋有内积$\langle\cdot, \cdot\rangle$以及相应范数$\|\cdot\|$的实Hilbert空间, $C$$Q$分别为$H_1$$H_2$的非空闭凸子集.设$f: C\times C\rightarrow{\Bbb R}^1$$h: Q\times Q\rightarrow{\Bbb R}^1$为两个二元函数.由何振华[1]提出的, 关于$f$$h$的分裂均衡问题是指寻求一点$x^*\in C$$Ax^*\in Q$, 使得

$ \begin{eqnarray} f(x^*, x)\geq0, \ \forall x\in C\ \mbox{且}\ h(Ax^*, y)\geq0, \ \forall y\in Q, \end{eqnarray} $ (1.1)

其中$A:H_1\rightarrow H_2$为一有界线性算子.显然分裂均衡问题包含两个均衡问题, 并且其中一个的解在该有界线性算子映照之下的像恰为另一个的解.由于许多物理, 优化以及经济问题均可归结为寻求解均衡问题的解[2-3], 其在应用数学领域中的地位尤为重要.以往, 某些均衡问题的公共解常被认为处于同一空间的同一子集中[4-6].然而, 一般而言, 不同均衡问题的解应属于不同空间的子集, 于是便产生分裂均衡问题.该问题的一个特殊情形便是分裂变分不等式问题[7].

例1.1[1]  设$H_1=H_2={\Bbb R}^1$, $C=[1, \infty)$$Q=(-\infty, -4]$.设$A(x)=-4x$, $x\in{\Bbb R}^1$, 则$A$为一有界线性算子.定义$g:C\times C\rightarrow{\Bbb R}^1$$h:Q\times Q\rightarrow{\Bbb R}^1$分别为$g(x, y)=y-x, \ h(u, v)=2(u-v)$.显然有$(EP)_g=\{1\}$且$A(1)=-4\in(EP)_h$, 其中$(EP)_f$表示关于二元函数$f$的均衡问题.于是$\Omega_{g, h}:=\{x^*\in(EP)_g:Ax^*\in(EP)_h\}\neq\emptyset$.

2012年, 何振华[1]构造迭代算法求解实Hilbert空间中的分裂均衡问题并获得弱收敛及强收敛定理.然而, 其研究结果局限于仅有有限多个均衡问题组成的情形.受此启发, 本文提出一类新的研究对象, 即:分裂均衡问题系统.设$\{f_i\}^{\infty}_{i=1}:C\times C\rightarrow{\Bbb R}^1$$\{h_i\}^{\infty}_{i=1}:Q\times Q\rightarrow{\Bbb R}^1$为两列二元函数, 关于$\{f_i\}$$\{h_i\}$的分裂均衡问题系统是指寻求一点$x^*\in C$$Ax^*\in Q$, 使得

$ \begin{eqnarray} f_i(x^*, x)\geq0, \ \forall x\in C\ \mbox{且}\ h_i(Ax^*, y)\geq0, \ \forall y\in Q, \ i=1, 2, \cdots, \end{eqnarray} $ (1.2)

其解集以$\Omega:=\bigcap\limits^{\infty}_{i=1}\Omega_{f_i, h_i}$表示, 其中$A:H_1\rightarrow H_2$为一有界线性算子.

例1.2  设$H_1={\Bbb R}^2$, $H_2={\Bbb R}^1$, 分别赋有标准范数$\|\cdot\|$$|\cdot|$; $C=\{(x_1, x_2)\in{\Bbb R}^2:x_2-x_1\geq1\}$$Q=[1, \infty)$.定义一列二元函数$\{f_i\}^{\infty}_{i=1}:C\times C\rightarrow{\Bbb R}^1$

$ {f_i}(x,y) = \left\{ {\begin{array}{*{20}{c}} {{{({x_1} - {x_2})}^i} + \sqrt[i]{{{y_2} - {y_1}}},\;\;\;i = 2k - 1,\;k = 1,2, \cdots ,}\\ {\sqrt[i]{{{y_2} - {y_1}}} - {{({x_1} - {x_2})}^i},\;\;i = 2k,\;\;\;\;\;\;k = 1,2, \cdots ,} \end{array}} \right. $

其中$x=(x_1, x_2)$, $y=(y_1, y_2)$.显然有$\bigcap\limits^{\infty}_{i=1}(EP)_{f_i}=\{(p_1, p_2)\in{\Bbb R}^2:p_2-p_1=1\}$.设$Ax=x_2-x_1$, $x=(x_1, x_2)\in{\Bbb R}^2$, 则$A$$H_1$$H_2$的有界线性算子且有$\|A\|=\sqrt{2}$.再定义一列二元函数$\{h_i\}^{\infty}_{i=1}:Q\times Q\rightarrow{\Bbb R}^1$

$ {h_i}(u,v) = \sqrt[i]{v} - \frac{u}{i},\;i = 1,2, \cdots . $

则显然有$\bigcap\limits^{\infty}_{i=1}(EP)_{h_i}=\{1\}$, 且对任一$x^*\in\bigcap\limits^{\infty}_{i=1}(EP)_{f_i}$, $Ax^*=1\in\bigcap\limits^{\infty}_{i=1}(EP)_{h_i}$, 即$\Omega\neq\emptyset$.

本文旨在求解一类关于两列二元函数的分裂均衡问题系统.该问题可转换为:以迭代算法逼近关于两族非线性算子的分裂公共不动点问题的解.文中于Hilbert空间背景下建立了强收敛定理.该结果推广了相关研究中均衡问题仅为有限多个的情形.

2 预备知识

为获得本文主要结果, 我们首先回顾如下定义, 记号以及结论.

$K$为实Hilbert空间$H$的一非空闭凸子集. $H$$K$上的度量投影$P_K : H \rightarrow K$定义为:对任一$x \in H$, 存在唯一点$z = P_K(x)$, 使得

$ ||x - z|| = \mathop {\inf }\limits_{y \in K} {\mkern 1mu} x - y = d(x,K). $

引理2.1  对任意的$x \in H$$z \in K$, 我们有

(1)   $z = P_K(x)$当且仅当如下关系式成立

$ \langle x - z,y - z\rangle \le 0,\;\forall y \in K. $

(2)

$ \langle {P_K}(x) - {P_K}(y),x - y\rangle \ge {P_K}(x) - {P_K}(y){^2},\;\forall x,y \in H. $

这表明$P_K : H\rightarrow K$是非扩张映射.

(3)

$ \|y-{{P}_{K}}(x){{\|}^{2}}+\|x-{{P}_{K}}(y){{\|}^{2}}\le \|x-y{{\|}^{2}},\ \forall x,y\in H. $

$E$为一Banach空间.称映射$T:E\rightarrow E$半闭于原点, 若对任一序列$\{x_n\}\subset E$满足$x_n\rightharpoonup x^*$$\|x_n-Tx_n\|\rightarrow0$, 则有$x^*=Tx^*$, 其中$x_n\rightharpoonup x^* $表示$\{x_n\}$弱收敛于$x^*$.

称Banach空间$E$满足Opial条件, 若对$E$中的任一序列$\{x_n\}$满足$x_n\rightharpoonup x^* $, 则有

$ \begin{eqnarray*} \liminf_{n\rightarrow\infty}\|x_n-x^*\|<\liminf_{n\rightarrow\infty}\|x_n-y\|, \ \forall y\in E, y\neq x^*. \end{eqnarray*} $

众所周知, 每一Hilbert空间均满足Opial条件.

引理2.2[8]  设$K$为实Hilbert空间$H$的一非空闭凸子集, $f: K\times K\rightarrow {\Bbb R}^1$为一满足如下条件的二元函数:

($A_1$)  $f(x, x)=0$;

($A_2$)  $f$是单调的, 即$f(x, y)+f(y, x)\leq0$;

($A_3$)  $\limsup\limits_{t\downarrow0}f(x+t(z-x), y)\leq f(x, y)$;

($A_4$)  映射$y\mapsto f(x, y)$是凸且下半连续的.

$r>0$$x\in H$.定义映射$T^{f}_r:H\rightarrow K$如下

$ \begin{eqnarray} T^{f}_r(x)=\left\{z\in K:f(z, y)+\frac{1}{r}\langle y-z, z-x\rangle\geq0, \forall y\in K\right\}. \end{eqnarray} $ (2.1)

于是有

(1)  $T^{f}_r$是单值的;

(2)  $T^{f}_r$是固定非扩张的, 即对任意的$x, y\in H$,

$ {{\left\| T_{r}^{f}(x)-T_{r}^{f}(y) \right\|}^{2}}\le \left\langle T_{r}^{f}(x)-T_{r}^{f}(y),x-y \right\rangle ; $

(3)  $F\left(T^{f}_r\right)=(EP)_{f}$, $\forall r>0$;

(4)  $(EP)_{f}$是闭且凸的.

引理2.3[9]  设$H$为一Hilbert空间.则对所有$x, y\in H$及满足$\sum\limits^{n}_{i=0}\alpha_i=1$$\alpha_i\in[0,1] $, $i = 0, 1, 2, \cdots, n$, 有如下等式成立

$ \begin{eqnarray} \left\|\sum^{n}_{i=0}\alpha_ix_i\right\|^2=\sum^{n}_{i=0}\alpha_i\left\|x_i\right\|^2-\sum_{0\leq i, j\leq n}\alpha_i\alpha_j\left\|x_i-x_j\right\|^2. \end{eqnarray} $ (2.2)

引理2.4  设$H$为一实Hilbert空间.对任意的$x, y\in H$, 有如下关系式成立

$ \|x+y{{\|}^{2}}\le \|y{{\|}^{2}}+2\langle x,x+y\rangle $

$ \|x-y{{\|}^{2}}=\|x{{\|}^{2}}+\|y{{\|}^{2}}-2\langle x,y\rangle . $

引理2.5[10]  正整数方程

$ \begin{eqnarray} n=i_n+\frac{(m_n-1)m_n}{2}, \ m_n\geq i_n, n=1, 2, 3, \cdots \end{eqnarray} $ (2.3)

的唯一解为

$ \begin{eqnarray} i_n=n-\frac{(m_n-1)m_n}{2}, \ m_n=-\left[\frac{1}{2}-\sqrt{2n+\frac{1}{4}}\right], \ n=1, 2, 3, \cdots, \end{eqnarray} $ (2.4)

其中$[x]$代表不超过$x$的最大整数.

引理2.6[11]  设$E$为一光滑, 严格凸且自反的Banach空间, $C$$E$的非空闭凸子集.我们有如下结论成立

$ \begin{eqnarray} \mbox{若}\ x\in E\ \mbox{且}\ z\in C, \ \mbox{则}\ z=\Pi_{C}x\Leftrightarrow\langle z-y, Jx-Jz\rangle\geq0, \ \forall y\in C, \end{eqnarray} $ (2.5)

其中$\Pi_{C}:E\rightarrow C$表示由下式定义的广义投影

$ {\Pi _C} = \arg \mathop {\inf }\limits_{y \in C} {\mkern 1mu} \left\{ {y{^2} - 2\langle y,Jx\rangle + x{^2}} \right\},\;\forall x \in E. $
3 主要结果

本节将求解满足条件$(A_1)-(A_4)$的分裂均衡问题系统.

定理3.1  设$C$$Q$分别为实Hilbert空间$H_1$$H_2$的非空闭凸子集.设$\{f_i\}:C\times C\rightarrow{\Bbb R}^1$$\{h_i\}:Q\times Q\rightarrow{\Bbb R}^1$为两列二元函数, 且有$\bigcap\limits^{\infty}_{i=1}(EP)_{f_i}\neq\emptyset$$\bigcap\limits^{\infty}_{i=1}(EP)_{h_i}\neq\emptyset$.设$A : H_1\rightarrow H_2$为一有界线性算子, $A^*$为其对偶算子.迭代序列$\{x_n\}$定义如下

$ \left\{ \begin{array}{*{35}{l}} {{x}_{1}}\in {{C}_{1}}=:C, \\ {{f}_{{{i}_{n}}}}({{u}_{n}},y)+\frac{1}{{{r}_{{{i}_{n}}}}}\langle y-{{u}_{n}},{{u}_{n}}-{{x}_{n}}\rangle \ge 0,\forall y\in C, \\ {{h}_{{{i}_{n}}}}({{w}_{n}},z)+\frac{1}{{{r}_{{{i}_{n}}}}}\langle z-{{w}_{n}},{{w}_{n}}-A{{u}_{n}}\rangle \ge 0,\forall z\in Q, \\ {{y}_{n}}={{P}_{C}}\left( {{u}_{n}}+\mu {{A}^{*}}({{w}_{n}}-A{{u}_{n}}) \right), \\ {{C}_{n+1}}=\{v\in {{C}_{n}}:\|{{y}_{n}}-v\|\le \|{{u}_{n}}-v\|\le \|{{x}_{n}}-v\|\}, \\ {{x}_{n+1}}={{P}_{{{C}_{n+1}}}}{{x}_{1}},\ \forall n\in \mathbb{N}, \\ \end{array} \right. $ (3.1)

其中$\{r_i\}$$(0, \infty)$中的一列实数, $\mu\in\big(0, \frac{1}{\|A^*\|^2}\big)$, $ $P_{C_{n+1}}$为$H_1$$C_{n+1}$上的投影算子; $i_n$为正整数方程$n=i_n+\frac{(m_n-1)m_n}{2}\ (m_n\geq i_n, n=1, 2, \cdots)$的解, 即对每一$n\geq1$, 存在唯一的$i_n$, 使得

$ {{i}_{1}}=1,{{i}_{2}}=1,{{i}_{3}}=2,{{i}_{4}}=1,{{i}_{5}}=2,{{i}_{6}}=3,{{i}_{7}}=1,{{i}_{8}}=2,{{i}_{9}}=3,{{i}_{10}}=4,{{i}_{11}}=1,\cdots . $

$\Omega:=\{x^*\in\bigcap\limits^{\infty}_{i=1}(EP)_{f_i}:Ax^*\in\bigcap\limits^{\infty}_{i=1}(EP)_{h_i}\}\neq\emptyset$, 则$\{x_n\}$强收敛于一点$x^*\in\Omega$.

  由引理2.2, 对所有$ n\in{\Bbb N}$, $u_n=T^{f_{i_n}}_{r_{i_n}}x_n$$w_n=T^{h_{i_n}}_{r_{i_n}}Au_n$, 其中, 对任一满足条件$(A_1)-(A_4)$的二元函数$\varphi$及任一常数$r>0$, $T^{\varphi}_{r}$按(2.1)式定义.先证对所有$n\in{\Bbb N}$, $\Omega\subset C_{n+1}$.由引理2.2和引理2.4可得

$ \|{{u}_{n}}-p{{\|}^{2}}={{\left\| T_{{{r}_{{{i}_{n}}}}}^{{{f}_{{{i}_{n}}}}}{{x}_{n}}-T_{{{r}_{{{i}_{n}}}}}^{{{f}_{{{i}_{n}}}}}p \right\|}^{2}} \\ \text{ }\le \left\langle T_{{{r}_{{{i}_{n}}}}}^{{{f}_{{{i}_{n}}}}}{{x}_{n}}-T_{{{r}_{{{i}_{n}}}}}^{{{f}_{{{i}_{n}}}}}p,{{x}_{n}}-p \right\rangle \\ \text{ }=\frac{1}{2}\left( \|{{u}_{n}}-p{{\|}^{2}}+\|{{x}_{n}}-p{{\|}^{2}}-\|{{u}_{n}}-{{x}_{n}}{{\|}^{2}} \right),\text{ } $

于是有

$ \begin{equation} \|u_n-p\|^2\leq\|x_n-p\|^2-\|u_n-x_n\|^2. \end{equation} $ (3.2)

又由引理2.2得到

$ \begin{equation} \|w_n-Ap\|^2=\left\|T^{h_{i_n}}_{r_{i_n}}Au_n-Ap\right\|^2\leq\|Au_n-Ap\|^2. \end{equation} $ (3.3)

由引理2.4和(3.3)式可得

$ \begin{eqnarray} && 2\mu\left\langle u_n-p, A^*\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\rangle \nonumber\\ &=&2\mu\left\langle A(u_n-p)+\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n-\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n, \left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\rangle\nonumber\\ &=&2\mu\left(\left\langle T^{h_{i_n}}_{r_{i_n}}Au_n-Ap, \left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\rangle-\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2\right)\nonumber\\ &=&2\mu\left(\frac{1}{2}\left\|T^{h_{i_n}}_{r_{i_n}}Au_n-Ap\right\|^2-\frac{1}{2}\|Au_n-Ap\|^2-\frac{1}{2}\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2\right)\nonumber\\ &\leq&-\mu\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2 \end{eqnarray} $ (3.4)

$ \begin{eqnarray} \left\|A^*\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2\leq\|A^*\|^2\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2. \end{eqnarray} $ (3.5)

由(3.1), (3.4)及(3.5)式可得

$ \begin{eqnarray} \|y_n-p\|^2&\leq&\left\|u_n+\mu A^*\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n-p\right\|^2\nonumber\\ &=&\|u_n-p\|^2+\left\|\mu A^*\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2+2\mu\left\langle u_n-p, A^*\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\rangle\nonumber\\ &\leq&\|u_n-p\|^2+\mu^2\|A^*\|^2\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2-\mu\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2\nonumber\\ &=&\|u_n-p\|^2-\mu\left(1-\mu\|A^*\|^2\right)\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2\nonumber\\ &\leq&\|x_n-p\|^2-\mu\left(1-\mu\|A^*\|^2\right)\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2. \end{eqnarray} $ (3.6)

注意到$\mu\in\left(0, \frac{1}{\|A^*\|^2}\right)$, 则有$\mu\big(1-\mu\|A^*\|^2\big)>0$.于是由(3.6)式得到

$ \begin{eqnarray} \|y_n-p\|\leq\|u_n-p\|\leq\|x_n-p\|, \end{eqnarray} $ (3.7)

这表明$p\in C_{n+1}$, 即对所有$n\in{\Bbb N}$, $\Omega\subset C_{n+1}$.

再证$C_{n+1}$为一闭凸集.其闭性无须赘述, 只证凸性.事实上, 对任意的$v_1, v_2\in C_{n+1}$$t\in(0, 1)$, 由引理2.3可得

$ \begin{eqnarray*} \|y_n-(tv_1+(1-t)v_2)\|^2&=&\|t(y_n-v_1)+(1-t)(y_n-v_2)\|^2 \nonumber\\ &=&t\|y_n-v_1\|^2+(1-t)\|y_n-v_2\|^2-t(1-t)\|v_1-v_2\|^2\nonumber\\ &\leq&t\|u_n-v_1\|^2+(1-t)\|u_n-v_2\|^2-t(1-t)\|v_1-v_2\|^2\nonumber\\ &=&\|u_n-(tv_1+(1-t)v_2)\|^2. \end{eqnarray*} $

类似有$ \|u_n-(tv_1+(1-t)v_2)\|^2\leq\|x_n-(tv_1+(1-t)v_2)\|^2$.这表明$tv_1+(1-t)v_2\in C_{n+1}$, 因此, 对所有$n\in{\Bbb N}$, $C_{n+1}$均为凸集.

现往证$x_n\rightarrow x^*\in H$ $(n\rightarrow\infty)$.根据引理2.2 (4), $\Omega$为一闭凸集, 于是存在唯一一点$\omega\in\Omega\subset C_{n}$使得$\omega=P_{\Omega}x_1$.由$x_{n}=P_{C_{n}}x_1$可得, 对所有$n\in{\Bbb N}$, $\|x_{n}-x_1\|\leq\|\omega-x_1\|$.这表明$\{x_n\}$有界, $\{u_n\}$$\{y_n\}$亦然.此外, 由$x_{n}=P_{C_{n}}x_1$$x_{n+1}\in C_{n+1}\subset C_{n}$可得

$ \begin{eqnarray} \|x_{n}-x_1\|\leq\|x_{n+1}-x_1\|, \ \forall n\in{\Bbb N}. \end{eqnarray} $ (3.8)

因此, 极限$\lim\limits_{n\rightarrow\infty}\|x_{n}-x_1\|$存在.由引理2.1 (3)及$x_{m}=P_{C_{m}}x_1\subset C_n$, 对任意的正整数$m$$n$, $m >n$, 我们有

$ \begin{eqnarray*} \|x_{n}-x_m\|^2+\|x_{1}-x_m\|^2&=&\left\|x_{n}-P_{C_{m}}x_1\right\|^2+\left\|x_{1}-P_{C_{m}}x_1\right\|^2\\ &\leq&\|x_{n}-x_1\|^2, \end{eqnarray*} $

亦即

$ \begin{eqnarray} \|x_{n}-x_m\|^2\leq\|x_{n}-x_1\|^2-\|x_{m}-x_1\|^2. \end{eqnarray} $ (3.9)

由极限$\lim\limits_{n\rightarrow\infty}\|x_{n}-x_1\|$的存在性可知$\{x_n\}$为一Cauchy列, 于是存在一点$x^*\in H$使得

$ \begin{eqnarray} x_{n}\rightarrow x^*\ (n\rightarrow\infty). \end{eqnarray} $ (3.10)

最后, 我们证明$x^*\in\Omega$.由(3.1)和(3.10)式可得

$ \begin{eqnarray} \|y_n-x_n\|&\leq&\|y_n-x_{n+1}\|+\|x_{n+1}-x_{n}\|\nonumber\\ &\leq&2\|x_{n+1}-x_{n}\|\rightarrow0\ (n\rightarrow\infty), \end{eqnarray}$ (3.11)
$ \begin{eqnarray} \|u_n-x_n\|&\leq&\|u_n-x_{n+1}\|+\|x_{n+1}-x_{n}\|\nonumber\\ &\leq&2\|x_{n+1}-x_{n}\|\rightarrow0\ (n\rightarrow\infty) \end{eqnarray} $ (3.12)

$ \begin{equation} \|y_n-u_n\|\leq\|y_n-x_{n}\|+\|x_{n}-u_{n}\|\rightarrow0\ (n\rightarrow\infty). \end{equation} $ (3.13)

又由(3.6)式得到

$ \begin{eqnarray} \left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|^2&\leq&\frac{1}{\mu\left(1-\mu\|A^*\|^2\right)}\left(\|x_n-p\|^2-\|y_{n}-p\|^2\right)\nonumber\\ &\leq&\frac{1}{\mu\left(1-\mu\|A^*\|^2\right)}\|x_n-y_n\|(\|x_n-p\|-\|y_{n}-p\|), \end{eqnarray} $ (3.14)

于是有

$ \begin{equation} \lim\limits_{n\rightarrow\infty}\left\|\left(T^{h_{i_n}}_{r_{i_n}}-I\right)Au_n\right\|=0. \end{equation} $ (3.15)

由(3.12)式可得

$ \begin{equation} \lim\limits_{n\rightarrow\infty}\left\|T^{f_{i_n}}_{r_{i_n}}x_n-x_n\right\|=\lim\limits_{n\rightarrow\infty}\|u_{n}-x_n\|=0. \end{equation} $ (3.16)

${\Bbb N}_i=\left\{k\in{\Bbb N}:k=i+\frac{(m-1)m}{2}, m\geq i, m\in{\Bbb N}\right\}$, $i\in{\Bbb N}$.例如, 根据引理2.5及${\Bbb N}_1$的定义.我们有$ {\Bbb N}_1=\{1, 2, 4, 7, 11, 16, \cdots\} $$ i_1=i_2=i_4=i_7=i_{11}=i_{16}=\cdots=1$.注意到只要$k\in {\Bbb N}_i$, 便有$T^{f_{i_n}}_{r_{i_n}}=T^{f_{i}}_{r_{i}}$.于是由(3.16)式得到

$ \begin{eqnarray} \lim\limits_{{\Bbb N}_i\ni k\rightarrow\infty}\left\|T^{f_{i}}_{r_{i}}x_k-x_k\right\|=0, \forall i\in{\Bbb N}. \end{eqnarray} $ (3.17)

由于$\{x_k\}_{k\in {\Bbb N}_i}$$\{x_n\}$的子列, 则由(3.10)式可得$x_k\rightarrow x^{*}$ $({\Bbb N}_i\ni k\rightarrow\infty)$.由(3.17)式及$T^{f_{i}}_{r_{i}}$的连续性立刻得到:对每一$i\in{\Bbb N}$, $x^{*}\in F\left(T^{f_{i}}_{r_{i}}\right)$, 于是有$x^{*}\in \bigcap\limits^{\infty}_{i=1}(EP)_{f_i}$.

根据(3.15)式, 且注意到只要$k\in {\Bbb N}_i$, 便有$T^{h_{i_n}}_{r_{i_n}}=T^{h_{i}}_{r_{i}}$, 我们亦有

$ \begin{eqnarray} \lim\limits_{{\Bbb N}_i\ni k\rightarrow\infty}\left\|T^{h_{i}}_{r_{i}}Au_k-Au_k\right\|=0, \forall i\in{\Bbb N}, \end{eqnarray} $ (3.18)

这意味着$Ax^{*}\in \bigcap\limits^{\infty}_{i=1}(EP)_{h_i}$, 于是有$x^{*}\in\Omega$.这显然是由于

$ \begin{eqnarray} \|u_n-x^*\|\leq\|u_n-x_{n}\|+\|x_{n}-x^*\|\rightarrow0\ (n\rightarrow\infty) \end{eqnarray} $ (3.19)

以及每一$T^{h_{i}}_{r_{i}}$$A$的连续性.证毕.

例3.2  设$H_1=H_2={\Bbb R}^1$且赋有标准范数$\|\cdot\|=|\cdot|$; $C=Q=[0,1]$.设$\{f_i\}:C\times C\rightarrow {\Bbb R}^1$$\{h_i\}:Q\times Q\rightarrow{\Bbb R}^1$为两列二元函数, 分别定义为$ f_i(x, y)=\frac{1}{{1+i}}(y-x), \ h_i(u, v)=\frac{i-1}{{i}}(uv-u^2)$.设$Ax=A^*x=\frac{x}{2}$ $(x\in C)$, $\mu=3\in\big(0, \frac{1}{\|A^*\|^2}\big)=(0, 4)$, 以及$r_i=i\ (i=1, 2, \cdots)$.于是迭代序列(3.1)可重写为

$ \left\{ \begin{array}{*{35}{l}} {{x}_{1}}\in {{C}_{1}}=:C, \\ \left\langle y-{{u}_{n}},{{u}_{n}}-\left( {{x}_{n}}-\frac{{{i}_{n}}}{1+{{i}_{n}}} \right) \right\rangle \ge 0,\forall y\in C, \\ \left\langle z-{{w}_{n}},{{w}_{n}}-\frac{{{u}_{n}}}{2{{i}_{n}}} \right\rangle \ge 0,\forall z\in Q, \\ {{y}_{n}}={{P}_{C}}\left( {{u}_{n}}+\frac{3}{2}\left( {{w}_{n}}-\frac{{{u}_{n}}}{2} \right) \right), \\ {{C}_{n+1}}=\{v\in {{C}_{n}}:\|{{y}_{n}}-v\|\le \|{{u}_{n}}-v\|\le \|{{x}_{n}}-v\|\}, \\ {{x}_{n+1}}={{P}_{{{C}_{n+1}}}}{{x}_{1}},\ \forall n\in \mathbb{N}, \\ \end{array} \right. $ (3.20)

其中$i_n$为正整数方程$n=i_n+\frac{(m_n-1)m_n}{2}\ (m_n\geq i_n, n=1, 2, \cdots)$的解.不难得出:对所有$n\in{\Bbb N}$, $C_{n+1}=\left[0, \frac{x_n+y_n+u_n}{3}\right]$.由${\Bbb R}^1$为Hilbert空间可知$J=I$$\Pi_C=P_C$.于是, 根据引理2.6, (3.20)式转化为

$ \left\{ \begin{array}{*{35}{l}} {{x}_{1}}\in {{C}_{1}}=:[0,1], \\ {{u}_{n}}={{P}_{C}}\left( {{x}_{n}}-\frac{{{i}_{n}}}{1+{{i}_{n}}} \right), \\ {{w}_{n}}={{P}_{C}}\left( \frac{{{u}_{n}}}{2{{i}_{n}}} \right), \\ {{y}_{n}}={{P}_{C}}\left( {{u}_{n}}+\frac{3}{2}\left( {{w}_{n}}-\frac{{{u}_{n}}}{2} \right) \right), \\ {{x}_{n+1}}=\frac{{{x}_{n}}+{{y}_{n}}+{{u}_{n}}}{3},\ \forall n\in \mathbb{N}. \\ \end{array} \right. $ (3.21)

显然, $\{f_i\}$$\{h_i\}$满足条件$(A_1)-(A_4)$且有$\Omega=\{0\}$.于是由定理3.1可知$\{x_n\}$强收敛到零点.利用MATLAB 7.10.0.499所得的数值实验结果表明, 当$x_1=1$时, $x_{10}, \ x_{15}$$x_{20}$的计算结果分别为0.00015241579, 0.00000062722547和2.5811748e-9.这一例子验证了本文所设算法用于解决分裂均衡问题系统的有效性.

参考文献
[1] He Z. The split equilibrium problem and its convergence algorithms. Journal of Inequalities and Applications , 2012, 2012 : 162. DOI:10.1186/1029-242X-2012-162
[2] Blum E, Oettli W. From optimization and variational inequalities to equilibrium problems. Math Stud , 1994, 63 : 123–145.
[3] Moudafi A, Théra M. Proximal and dynamical approaches to equilibrium problems//Lecture Notes in Economics and Mathematical Systems. Berlin: Springer , 1999, 477 : 187–201.
[4] He Z, Du W S. Strong convergence theorems for equilibrium problems and fixed point problems: A new iterative method, some comments and applications. Fixed Point Theory and Applications , 2011, 2011 : 33. DOI:10.1186/1687-1812-2011-33
[5] Colao V, Acedo G L, Marino G. An implicit method for finding common solutions of variational inequalities and systems of equilibrium problems and fixed points of infinite family of nonexpansive mappings. Nonlinear Anal , 2009, 71 : 2708–2715. DOI:10.1016/j.na.2009.01.115
[6] Saeidi S. Iterative algorithms for finding common solutions of variational inequalities and systems of equilibrium problems and fixed points of families and semigroups of nonexpansive mappings. Nonlinear Anal , 2009, 70 : 4195–4208. DOI:10.1016/j.na.2008.09.009
[7] Censor Y, Gibali A, Reich S. Algorithms for the Split Variational Inequality Problem. Numerical Algor , 2012, 59(2) : 301–323. DOI:10.1007/s11075-011-9490-5
[8] Combettes P L, Hirstoaga A. Equilibrium programming in Hilbert spaces. J. Nonlinear Convex Anal , 2005, 6 : 117–136.
[9] Zegeye H, Shahzad N. Convergence of Mann's type iteration method for generalized asymptotically nonexpansive mappings. Comput Math Appl , 2011, 62 : 4007–1014. DOI:10.1016/j.camwa.2011.09.018
[10] Deng W Q, Bai P. An implicit iteration process for common fixed points of two infinite families of asymptotically nonexpansive mappings in Banach spaces. J Appl Math , 2013, Art ID : 602582.
[11] Alber Y I. Metric and generalized projection operators in Banach spaces: properties and applications//Kartosator A G, ed. Theory and Applications of Nonlinear Operators of Accretive and monotone type. New York: Marcel Dekker, 1996: 15-50