数学杂志  2015, Vol. 35 Issue (1): 123-130   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
唐艳
非扩张半群的变分不等式的不动点解的粘性逼近方法
唐艳    
重庆工商大学数学与统计学院, 重庆 400067
摘要:本文研究了非扩张半群的变分不等式的不动点解的迭代算法.利用变分不等式与不动点问题的解的关系, 结合粘性逼近方法, 建立了非扩张半群的不动点的两步迭代格式, 证明了该方法所得到的迭代序列在一定条件下的强收敛性, 并收敛于某变分不等式的唯一解.
关键词非扩张    不动点    粘性逼近方法    强收敛    变分不等式    
VISCOSITY APPROXIMATION METHODS FOR THE FIXED-POINT SOLUTIONS OF VARIATIONAL INEQUALITIES FOR NONEXPANSIVE SEMIGROUPS
TANG Yan    
College of Mathematics and Statistics, Chongqing Technology and Business University, Chongqing 400067, China
Abstract: The iterative scheme of fixed point solutions to variational inequalities of nonexpansive semigroups is researched in the real Banach space with a uniformly Gateaux differentiable normin this paper. Utilizing the relationship between the variational inequalities and fixed point solution, together with the approach of viscosity approximation, we establish the two-step iterative format of fixed point concerning the nonexpansive semigroups and prove the iterative sequences obtained by this approach possess strong convergence under certain conditions and will strongly converge to the unique solution of certain variational inequality.
Key words: nonexpansive     fixed point     viscosity approximation method     strong convergence     variational inequality    
1 引言

$E$为一实Banach空间, $C$$E$的一个非空闭凸子集.若映射$T:C\rightarrow C$满足条件

$\begin{equation*} \|Tx-Ty\|\leq {\|x-y\|}, \forall x,y\in{C}, \end{equation*}$

则称$T$为非扩张映射.记$F(T)$$T$的不动点集, 即$F(T)=\{x\in{C}:Tx=x\}$.

$f:C\rightarrow C$, 若存在一个实数$\rho\in(0,1)$并且满足条件

$\begin{equation*} \|f(x)-f(y)\|\leq {\rho\|x-y\|}, \forall x,y\in{C}, \end{equation*}$

则称$f$是具有压缩系数$\rho$的压缩映射.

$C$是Hilbert空间中的非空闭凸子集, 参数族$\Gamma=\{T(s):s\in{\mathbb{R}^+}\}$被称为非扩张半群, 若满足以下条件:

(ⅰ) $T(0)x=x$, $\forall x\in{C}$;

(ⅱ) $T(s+t)=T(s)T(t)$, $\forall s,t \in{\mathbb{R}^+}$;

(ⅲ) $\|T(s){x}- T(s){y}\|\leq {\|x-y\|}$, $\forall x,y\in{C},s\geq 0$;

(ⅳ) $\forall x\in{C}$, 映射$T(\cdot)x:\mathbb{R}^+\rightarrow C$是连续的.

$F(\Gamma)$表示$\Gamma$的不动点集, 即$F(\Gamma)=\{x\in C:T(s)x=x,0\leq s<\infty\}$.

粘性逼近方法[1-2, 4-7, 9-12]是一种非常重要的方法, 因为它们经常被应用于凸最优化, 线性规划, 单调包含以及椭圆微分方程等等.

1967年Browder[1]首先在Banach空间中固定$u$, 讨论了迭代格式

$\begin{align} x_t=tu+(1-t)Tx_t \end{align}$ (1.1)

在一定条件下的强收敛性.

1976年, Brezis [2]在Hilbert空间中讨论了非扩张半群的迭代格式

$\begin{align} x_t=\frac{1}{t}\int _0^t T(s)x_tds,t\in(0,1), \end{align}$ (1.2)

不过仅仅得到了弱收敛性.

2008年Somyot [5]等在Hilbert空间中讨论了非扩张半群不动点的如下迭代格式:

$\begin{align} x_t=tf(x)+(1-t)\frac{1}{\lambda_t}\int _0^{\lambda_t} T(s)xds, \end{align}$ (1.3)

并在适当条件下证明了由(1.3) 式得到的序列的强收敛性.

2012年Rabian [6]等在上述文献的基础上, 在Banach空间中讨论了非扩张半群的迭代算法

$\begin{equation} \label{} \begin{cases} y_n=\varepsilon_nx_n+(1-\varepsilon_n)T(t_n)x_n, \\ x_{n+1}=\alpha_nf(x_n)+(1-\alpha_n)y_n. \end{cases} \end{equation}$ (1.4)

在此基础上, 本文将在具有一致Gateaux可微范数的Banach空间$C$中, 建立了一个逼近非扩张半群公共不动点的粘性逼近方法

$\begin{equation} \label{} \begin{cases} x_0=x\in{C}, \\ y_n=\varepsilon_nx_n+(1-\varepsilon_n)T(s)x_n, \\ x_{n+1}=\alpha_nf(x_n)+\beta_nx_n+\gamma_n \frac{1}{s_n}\int _0^{s_n} T(s)y_nds, \end{cases} \end{equation}$ (1.5)

其中$\{\alpha_n\}$, $\{\beta_n\}$, $\{\gamma_n\}$, $\{\varepsilon_n\}$$(0,1)$中的实数序列, 并证明该迭代方法强收敛到某类变分不等式问题的唯一解.

2 预备知识

$E$为一实Banach空间, $E^*$$E$的对偶空间, $\langle\cdot,\cdot\rangle$表示广义对偶对, 称$J:E\rightarrow{2^{E^*}}$为正规对偶映像, 如果

$\begin{equation*} J_x=\{f^*\in {E^*}:\langle x,f^*\rangle=\|x\|^2=\|f^*\|^2\},\forall x\in {E}. \end{equation*}$

今后均用$j$表示单值赋范对偶映射.

$S=\{x\in{E}:\|x\|=1\}$$E$的单位球面, 对任意的$x,y\in{S}$, $\lim\limits_{t\rightarrow 0}\frac{\|x+ty\|-\|x\|}t$一致存在, 则称$E$的范数是一致Gateaux可微的.

$C$$E$的非空闭凸子集.参数族$\Gamma=\{T(s):s\in{\mathbb{R}^+}\}$为非扩张半群, $f:C\rightarrow C$是具有压缩系数$\rho$的压缩映射.

定义

$\begin{align} T_{t,n}x=\frac{(1-\alpha_n)t}{\gamma_n+t\beta_n}f(x) +\frac{(1-t)\gamma_n}{\gamma_n+t\beta_n}\frac{1}{\omega_t}\int _0^{\omega_t} T(s)xds, \end{align}$ (2.1)

其中$t\in(0,1)$, $n\geq 0$, $\lim\limits_{t\rightarrow 0}\omega_t=\infty$.

引理1 [2, 5]$C$为实Banach空间$E$的非空闭凸子集, $T:C\rightarrow C$为非扩张映射, $F(T)\neq\varnothing.$对任意固定的$u\in E$以及$t\in(0,1)$, 则由$C\ni x\mapsto tu+(1-t)Tx$所定义的压缩算子的唯一不动点$z_t$, 在$t\rightarrow 0$时强收敛于$p\in F(T)$, 其中$p$是与$u$的距离最近的一点.

引理2 [3]$H$为一个Hilbert空间, $C$为实$H$的非空闭凸子集, $f:C\rightarrow C$是具有压缩系数$\rho$的压缩映射, 则\begin{equation*} \langle x-y, (I-f)y\rangle\geq (1-\rho)\|x-y\|^2, x, y\in C. \end{equation*}

引理3 [5]$E$为一实Banach空间, $E^*$$E$的对偶空间, $J:E\rightarrow{2^{E^*}}$为正规对偶映像, 则对任意的$x,y\in E$, 有

$\begin{equation*} \|x+y\|^2\leq \|x\|^2+2 \langle y, j(x+y)\rangle, \forall j(x+y)\in {J(x+y)}. \end{equation*}$

引理4 [6-8]$\{a_n\}$, $\{b_n\}$, $\{c_n\}$均为非负实数列, 对任意$\lambda_n\in [0,1]$, 若存在正整数$N$使得\begin{equation*} a_{n+1}\leq(1-\lambda_n)a_n+b_n+c_n, \forall n\geq N, \end{equation*}其中$\sum\limits_{n=1}^{\infty}\lambda_n=\infty$, $b_n=o(\lambda_n)$, $\sum\limits_{n=1}^{\infty}c_n<\infty$, 则$\lim\limits_{n\rightarrow \infty}a_n= 0$.

3 主要结果

定理1 设$C$为Banach空间$E$的一个非空闭凸子集, 参数族$\Gamma=\{T(s):s\in{\mathbb{R}^+}\}$为非扩张半群, $f:C\rightarrow C$是具有压缩系数$\rho$的压缩映射.若算子$T_{t,n}$由(2.1) 式所定义, 则对任意的$t\in(0,1)$, $T_{t,n}$是压缩的.

 设$\forall x,y\in C$, 由$f$的压缩性以及$T$的非扩张性, 可得

$\begin{align*} & \|T_{t,n}x-T_{t,n}y\|\\ = &\|\frac{(1-\alpha_n)t}{\gamma_n+t\beta_n}f(x)+\frac{(1-t)\gamma_n}{\gamma_n+t\beta_n}\frac{1}{\omega_t}\int _0^{\omega_t} T(s)xds-\frac{(1-\alpha_n)t}{\gamma_n+t\beta_n}f(y)-\frac{(1-t)\gamma_n}{\gamma_n+t\beta_n}\frac{1}{\omega_t}\int _0^{\omega_t} T(s)yds\| \notag \\ \leq& \frac{(1-\alpha_n)t}{\gamma_n+t\beta_n}\|f(x)-f(y)\|+\frac{(1-t)\gamma_n}{\gamma_n+t\beta_n}\frac{1}{\omega_t}\int _0^{\omega_t}\|T(s)x-T(s)y\|ds \notag \\ \leq & \frac{\rho(1-\alpha_n)t+(1-t)\gamma_n}{\gamma_n+t\beta_n}\|x-y\| \notag \\ \leq& \frac{\gamma_n+t\rho\beta_n}{\gamma_n+t\beta_n}\|x-y\|. \end{align*}$

因为$\rho\in(0,1)$, 所以$0<\frac{\gamma_n+t\rho\beta_n}{\gamma_n+t\beta_n}<1$, 即$T_{t,n}$为压缩映射.由Banach压缩映象原理可知, $T_{t,n}$存在唯一的不动点, 记作$p_{t,n}$, 即

$\begin{equation} p_{t,n}=T_{t,n}p_{t,n}=\frac{(1-\alpha_n)t}{\gamma_n+t\beta_n}f(p_{t,n})+\frac{(1-t)\gamma_n}{\gamma_n+t\beta_n}\frac{1}{\omega_t}\int _0^{\omega_t}T(s)p_{t,n}\mbox{ds}. \end{equation}$ (3.1)

若固定$n$, 由引理1以及文献[5]可得$\lim\limits_{t\rightarrow 0}p_{t,n}=p_n\in F(\Gamma)$, 又$p_n$是与$f(p_n)$距离最近的点, 可记$p_n=\tilde{x}\in F(\Gamma)$, 所以$\lim\limits_{t\rightarrow 0}p_{t,n}=\tilde{x}\in F(\Gamma)$.

定理2 设$E$为具有一致Gateaux可微范数的实Banach空间, $C$$E$的非空闭凸子集, $\Gamma=\{T(s):s\in{\mathbb{R}^+}\}$为非扩张半群, $F(\Gamma)\neq\varnothing$, 算子$T_{t,n}$由(2.1) 式所定义.设$f:C\rightarrow C$是具有压缩系数$\rho$的压缩映射.序列$\{x_n\}$$\{y_n\}$由(1.5) 式所定义, 其中实数序列$\{\alpha_n\}$, $\{\beta_n\}$, $\{\gamma_n\}$, $\{\varepsilon_n\}$$(0,1)$中的实数序列, $n\geq 0$, $\lim\limits_{n\rightarrow \infty}s_n=\infty$, 且满足下列条件:

(ⅰ) $\alpha_n+\beta_n+\gamma_n=1$, $n\geq 0$;

(ⅱ) $\lim\limits_{n\rightarrow \infty}\alpha_n= 0$, $\sum\limits_{n=1}^{\infty}\alpha_n= \infty$;

(ⅲ) $\lim\limits_{n\rightarrow \infty}\gamma_n= 0$.

则序列$\{x_n\}$有界, 且强收敛于变分不等式$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0$的唯一解$\tilde{x}$, 其中$x,\tilde{x}\in F(\Gamma)$.

 设$p\in F(\Gamma)$, 由$T$的非扩张性得\begin{equation*} \|y_n-p\|=\|\varepsilon_nx_n+(1-\varepsilon_n)T(s)x_n-p\|\leq \|x_n-p\|. \end{equation*}

对于$\forall n \geq 0$, 由$T$的非扩张性和$f$的压缩性, 有

$\begin{align*} \|x_{n+1}-p\|&=\|\alpha_nf(x_n)+\beta_nx_n+\gamma_n\frac{1}{s_n}\int _0^{s_n}T(s)y_nds-p\| \notag \\ &\leq \alpha_n\|f(x_n)-p\|+\beta_n\|x_n-p\|+\gamma_n\|\frac{1}{s_n}\int _0^{s_n}T(s)y_nds-p\| \notag \\ &\leq \rho\alpha_n\|x_n-p\|+\alpha_n\|f(p)-p\|+(1-\alpha_n)\|x_n-p\| \notag \\ &=(1-(1-\rho)\alpha_n)\|x_n-p\|+\alpha_n\|f(p)-p\| \notag \\ &\leq \max\{\|x_0-p\|,\frac{f(p)-p}{1-\rho}\}, \end{align*}$

所以$\{x_n\}$有界, 进一步可得$\{y_n\}$, $\{f(x_n)\}$均有界.

另一方面, 令$\tilde{x}\in F(\Gamma)$, 由引理3可知

$\begin{align} & \|x_{n+1}-\tilde{x}\|^2=\|\alpha_nf(x_n)+\beta_nx_n+\gamma_n\frac{1}{s_n}\int _0^{s_n}T(s)y_nds-\tilde{x}\|^2 \notag \\ \leq& \|\gamma_n(\frac{1}{s_n}\int _0^{s_n}T(s)y_nds-\tilde{x})+\beta_n(x_n-\tilde{x})\|^2+2\alpha_n \langle f(x_n)-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle \notag \\ \leq& [\gamma_n\|\frac{1}{s_n}\int _0^{s_n}T(s)y_nds-\tilde{x}\|+\beta_n\|x_n-\tilde{x}\|]^2 +2\alpha_n \langle f(x_n)-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle \notag \\ \leq& (1-\alpha_n)^2\|x_n-\tilde{x}\|^2 +2\alpha_n \langle f(x_n)-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle \notag \\ \leq & (1-\alpha_n)^2\|x_n-\tilde{x}\|^2 +2\alpha_n\|f(x_n)-f(\tilde{x})\|\|x_{n+1}-\tilde{x}\|+2\alpha_n \langle f(\tilde{x})-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle \notag \\ \leq& (1-\alpha_n)^2\|x_n-\tilde{x}\|^2 +\rho\alpha_n(\|x_n-\tilde{x}\|^2+\|x_{n+1}-\tilde{x}\|^2)+2\alpha_n \langle f(\tilde{x})-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle . \end{align}$ (3.2)

若记$M=\sup\|x_n-\tilde{x}\|^2$, 整理(3.2) 式得

$\begin{align*} \|x_{n+1}-\tilde{x}\|^2 &\leq[1-\frac{2(1-\rho)}{1-\rho\alpha_n}\alpha_n]\|x_n-\tilde{x}\|^2+\frac{2\alpha_n}{1-\rho\alpha_n}\langle f(\tilde{x})-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle +\frac{\alpha_n^2}{1-\rho\alpha_n}M. \end{align*}$

由定理1, 不妨设$p_{t,n}$$T_{t,n}$的唯一不动点, 将(2.1) 式另记为

$\begin{align} p_{t,n}=tf(p_{t,n})+\frac{1-t}{1-\alpha_n}[\gamma_n\frac{1}{\omega_t}\int _0^{\omega_t}T(s)p_{t,n}ds+\beta_np_{t,n}]. \end{align}$ (3.3)

于是

$\begin{align*} x_n-p_{t,n}=& t(x_n-f(p_{t,n}))+(1-t)[\frac{\beta_n}{1-\alpha_n}(x_n-p_{t,n})\\ &+\frac{\gamma_n}{1-\alpha_n}(x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)p_{t,n}ds)]. \end{align*}$

所以

$\begin{align} & \|x_n-p_{t,n}\|^2 \leq(1-t)^2\|\frac{\beta_n}{1-\alpha_n}(x_n-p_{t,n})\notag\\ & +\frac{\gamma_n}{1-\alpha_n}(x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)p_{t,n}ds)\|^2+2t\langle x_n-f(p_{t,n}),j(x_n-p_{t,n})\rangle \notag \\ \leq &(1-t)^2[\frac{\beta_n}{1-\alpha_n}\|x_n-p_{t,n}\|+ \frac{\gamma_n}{1-\alpha_n}\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\| \notag \\ &+ \frac{\gamma_n}{1-\alpha_n}\|\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)p_{t,n}ds\|]^2 + 2t\langle x_n-f(p_{t,n}),j(x_n-p_{t,n})\rangle \notag \\ \leq& (1-t)^2[\|x_n-p_{t,n}\|+\frac{\gamma_n}{1-\alpha_n}\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\|]^2+2t\langle x_n-p_{t,n},j(x_n-p_{t,n}) \rangle \notag \\ &+2t\langle p_{t,n}-f(p_{t,n}),j(x_n-p_{t,n})\rangle \notag \\ \leq& (1-t)^2[\|x_n-p_{t,n}\|+\frac{\gamma_n}{1-\alpha_n}\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\|]^2 \notag \\ &+2t\|x_n-p_{t,n}\|^2+2t\langle p_{t,n}-f(p_{t,n}),j(x_n-p_{t,n})\rangle \notag \\ &=(1+t^2)\|x_n-p_{t,n}\|+(1-t)^2[\frac{\gamma_n}{1-\alpha_n}\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\|]^2 \notag \\ &+2(1-t)^2 \frac{\gamma_n}{1-\alpha_n}\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\|\|x_n-p_{t,n}\|+2t\langle p_{t,n}-f(p_{t,n}),j(x_n-p_{t,n})\rangle \notag \\ \leq& (1+t^2)\|x_n-p_{t,n}\|^2+\frac{\gamma_n}{1-\alpha_n}M_1+2t\langle p_{t,n}-f(p_{t,n}),j(x_n-p_{t,n})\rangle, \end{align}$ (3.4)

其中

$M_1=\sup\{\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\|^2+2\|x_n-\frac{1}{\omega_t}\int _0^{\omega_t}T(s)x_nds\|\|x_n-p_{t,n}\|,n\geq 0\}.$ (3.5)

整理(3.4) 式

$\begin{align} \langle f(p_{t,n})-p_{t,n},j(x_n-p_{t,n})\rangle\leq \frac{t}{2}M_2+ \frac{\gamma_n}{2t(1-\alpha_n)}M_1, \end{align}$

其中$M_2\geq \sup\{\|p_{t,n}-x_n\|^2,0<t<1, n\geq0\}$.由(3.5) 式并注意条件(ⅲ), 可得

$\begin{align} \lim\limits_{t\rightarrow 0}\sup \lim\limits_{n\rightarrow \infty} \sup \langle f(p_{t,n})-p_{t,n},j(x_n-p_{t,n})\rangle \leq 0. \end{align}$ (3.6)

又因为$\lim\limits_{t\rightarrow 0}p_{t,n}=\tilde{x}$, 所以

$ \lim\limits_{n\rightarrow \infty} \sup \langle f(\tilde{x})-\tilde{x},j(x_n-\tilde{x})\rangle \leq 0, $

所以

$\begin{align} \|x_{n+1}-\tilde{x}\|^2 \leq& [1-\frac{2(1-\rho)}{1-\rho\alpha_n}\alpha_n]\|x_n-\tilde{x}\|^2\notag\\ &+\frac{2(1-\rho)\alpha_n}{1-\rho\alpha_n}[\frac{1}{1-\rho}\langle f(\tilde{x})-\tilde{x},j(x_{n+1}-\tilde{x}) \rangle +\frac{\alpha_n}{2(1-\rho)}M]. \end{align}$ (3.7)

$\lambda_n=\frac{2(1-\rho)}{1-\rho\alpha_n}\alpha_n$, 则由(3.6), (3.7) 式和引理4, 可知

$\lim\limits_{n\rightarrow \infty}\|x_n-\tilde{x}\|=0,$

$\{x_n\}$强收敛于$\tilde{x}\in F(\Gamma)$.

下面说明$\tilde{x}$是变分不等式$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0,\tilde{x}\in F(\Gamma)$的唯一解.

根据式(3.1) 式可知

$\begin{align*} (I-f)p_{t,n}&=[1-\frac{\gamma_n+t\beta_n}{(1-\alpha_n)t}]p_{t,n}+\frac{(1-t)\gamma_n}{(1-\alpha_n)t}\frac{1}{\omega_t}\int_0^{\omega_t}T(s)p_{t,n}ds \notag \\ &=\frac{(t-1)\gamma_n}{(1-\alpha_n)t}(p_{t,n}-\frac{1}{\omega_t}\int_0^{\omega_t}T(s)p_{t,n} ds), \end{align*}$

$\forall p\in F(\Gamma)$, 根据引理2, 有

$\begin{align*} \langle(I-f)p_{t,n},p_{t,n}-p\rangle&=\langle\frac{(t-1)\gamma_n}{(1-\alpha_n)t}(p_{t,n}-\frac{1}{\omega_t}\int_0^{\omega_t}T(s)p_{t,n}ds), p_{t,n}-p\rangle \notag \\ &=\frac{(t-1)\gamma_n}{(1-\alpha_n)t}\langle\ \frac{1}{\omega_t}\int_0^{\omega_t}(p_{t,n}-T(s)p_{t,n})ds-\frac{1}{\omega_t}\int_0^{\omega_t}(p-T(s)p)ds,p_{t,n}-p\rangle \notag \\ &=\frac{(t-1)\gamma_n}{(1-\alpha_n)t}\frac{1}{\omega_t}\langle\int_0^{\omega_t}[(I-T(s))p_{t,n}-(I-T(s))p]ds,p_{t,n}-p\rangle \notag \\ &=\frac{(t-1)\gamma_n}{(1-\alpha_n)t}\frac{1}{\omega_t}\int_0^{\omega_t}\langle[(I-T(s))p_{t,n}-(I-T(s))p],p_{t,n}-p\rangle ds \leq 0, \end{align*}$

所以

$\langle(I-f)p_{t,n},p-p_{t,n}\rangle\geq 0,\forall p\in F(\Gamma).$

由于$\lim\limits_{t\rightarrow 0}p_{t,n}=\tilde{x}$, 所以

$\langle(I-f)\tilde{x},p-\tilde{x}\rangle\geq 0, $

因此$\tilde{x}$$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0,x\in F(\Gamma)$的解.

若另外有一个$\acute{x}\in F(\Gamma)$也是变分不等式$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0$的解, 则有

$\langle(I-f)\tilde{x},\acute{x}-\tilde{x}\rangle\geq 0$

以及$\langle(I-f)\acute{x},\tilde{x}-\acute{x}\rangle\geq 0$.两式相加得

$\begin{align*} (1-\rho)\|\tilde{x}-\acute{x}\|^2\leq \langle(I-f)\tilde{x}-(I-f)\acute{x},\tilde{x}-\acute{x}\rangle\leq 0, \end{align*}$

所以$\tilde{x}=\acute{x}$.因此$\tilde{x}$$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0$的唯一解.

定理3 设$E$为具有一致Gateaux可微范数的实Banach空间, $C$$E$的非空闭凸子集. $\Gamma=\{T(s):s\in{\mathbb{R}^+}\}$为非扩张半群, $F(\Gamma)\neq\varnothing$.设$f:C\rightarrow C$是具有压缩系数$\rho$的压缩映射, $T_{t,n}$为由(2.1) 式所定义的压缩映射.如果定义序列

$\begin{align} x_{n+1}=\alpha_nf(x_n)+\beta_nx_n+\gamma_n \frac{1}{s_n}\int _0^{s_n} T(s)x_nds, \end{align}$ (3.8)

其中实数序列$\{\alpha_n\}$, $\{\beta_n\}$, $\{\gamma_n\}$$(0,1)$中的实数序列,

$\alpha_n+\beta_n+\gamma_n=1, \lim\limits_{n\rightarrow \infty}\alpha_n= 0, \sum\limits_{n=1}^{\infty}\alpha_n= \infty, \lim\limits_{n\rightarrow \infty}s_n=\infty,$

则序列$\{x_n\}$有界, 且强收敛于变分不等式$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0$的唯一解$\tilde{x}$, 其中$x,\tilde{x}\in F(\Gamma)$.

 当$\varepsilon_n=1$时, 粘性迭代格式(1.5) 便退化为(3.8) 式, 由定理2类似可证.

定理4 设$E$为具有一致Gateaux可微范数的实Banach空间, $C$$E$的非空闭凸子集. $\Gamma=\{T(s):s\in{\mathbb{R}^+}\}$为非扩张半群, $F(\Gamma)\neq\varnothing$.设$f:C\rightarrow C$是具有压缩系数$\rho$的压缩映射, $T_{t,n}$为由(2.1) 式所定义的压缩映射.如果定义序列

$\begin{equation} \label{} \begin{cases} y_n=\varepsilon_nx_n+(1-\varepsilon_n)T(s)x_n, \\ x_{n+1}=\alpha_nf(x_n)+(1-\alpha_n)T(t)y_n, \end{cases} \end{equation}$ (3.9)

其中实数序列$\{\varepsilon_n\}$, $\{\alpha_n\}$$(0,1)$中的实数序列, $\lim\limits_{n\rightarrow \infty}\alpha_n= 0$, $\sum\limits_{n=1}^{\infty}\alpha_n= \infty$, $n\geq 0$, 则序列$\{x_n\}$有界, 且强收敛于变分不等式$\langle(I-f)\tilde{x},x-\tilde{x}\rangle\geq 0$的唯一解$\tilde{x}$, 其中$x,\tilde{x}\in F(\Gamma)$.

 当$\beta_n=0$时, 粘性迭代格式(1.5) 便退化为(3.9) 式, 由定理2类似可证.事实上, 该定理是文献[6]的推广.因为若$T=I$, 格式(3.9) 式便可成为(1.4) 式.

参考文献
[1] Browder F E. Convergence of approximates to fixed points of nonexpansive nonlinear mappings in Banach spaces[J]. Arch. Ration. Mech. Anal., 1967, 24: 82–90.
[2] Bailon J B, Brezis H. Une Remarque sur le comportement asymptotique des semigroupes non lineaires[J]. Houston J. Math., 1976, 2: 5–7.
[3] Reich S. Strong convergence theorems for resolvents of accretive operators in Banach spaces[J]. Math. Anal. App., 1980, 75: 287–292. DOI:10.1016/0022-247X(80)90323-6
[4] Chang S S. On Chidume's open questions andapproximation solutions of multivalued strongly accretive mappings equations in Banach spaces[J]. J. Math. Anal. Appl., 1997, 216: 94–111. DOI:10.1006/jmaa.1997.5661
[5] Somyot, Plubtieng, Rattanaporn Punpaeng. Fixed-point solutions of variational inequalities for nonexpansive semigroups in Hilbert spaces[J]. Math. Computer Modelling, 2008, 48: 279–286. DOI:10.1016/j.mcm.2007.10.002
[6] Rabian Wangkeeree, Pakkapon Preechasilp. Modified noor iterations for nonexpansive semigroups with generalized contraction in Banach spaces[J]. J. Ineq. Appl., 2012.
[7] 田有先, 陈六新. 有限一致拟-李卜希兹映象族公共不动点的逼近[J]. 西南大学学报(自然科学版), 2009, 31(4): 25–29.
[8] Reich S. On the Aymptotic behavior of nonlinear semigroups and the range of accretive operators[J]. J. Math. Anal. Appl., 1981, 79: 113–126. DOI:10.1016/0022-247X(81)90013-5
[9] Liu L S. Ishikawa and Mann iterative processes with errors for nonlinear strongly accretive mappings in Banach space[J]. J. Math. Anal. Appl., 1995, 194: 114–125. DOI:10.1006/jmaa.1995.1289
[10] 唐艳. Banach空间中非扩张映射不动点的粘性逼近方法[J]. 数学杂志, 2013, 33(1): 99–104.
[11] Wen DaoJun, Chen YiAn. Strong convergence of modified general iterative method for generalized equilibrium problems and fixed point problems of $k$-strict pseudo-contractions[J]. Fixed Point Theory Appl., 2012.
[12] 唐艳, 闻道君. Banach空间中渐近非扩张连续半群不动点的粘性逼近[J]. 数学的实践与认识, 2012, 42(9): 225–230.