Next Article in Journal
Digital Supply Chain through Dynamic Inventory and Smart Contracts
Previous Article in Journal
Some Bicyclic Graphs Having 2 as Their Laplacian Eigenvalues
Article

Convergence Theorems of Variational Inequality for Asymptotically Nonexpansive Nonself Mapping in CAT(0) Spaces

Graduate School of Education, Mathematics Education, Kyungnam University, Changwon, Gyeongnam 51767, Korea
Mathematics 2019, 7(12), 1234; https://doi.org/10.3390/math7121234
Received: 8 November 2019 / Revised: 5 December 2019 / Accepted: 9 December 2019 / Published: 12 December 2019

Abstract

The aim of this manuscript is to get the strong convergence theorems of the Moudafi’s viscosity approximation methods for an asymptotically nonexpansive nonself mapping in C A T ( 0 ) spaces.
Keywords: geodesic triangle; CAT(0) space; quasilinearization; retraction; asymptotically nonexpansive nonself mapping; viscosity approximation method; Δ-convergence; variational inequality; fixed point geodesic triangle; CAT(0) space; quasilinearization; retraction; asymptotically nonexpansive nonself mapping; viscosity approximation method; Δ-convergence; variational inequality; fixed point

1. Introduction

Let ( X , d ) be a metric space. A geodesic path joining p 1 X to p 2 X (or, a geodesic from p 1 to p 2 ) is a mapping g from a closed interval [ 0 , l ] R to X such that g ( 0 ) = p 1 , g ( l ) = p 2 , and
d ( g ( t ) , g ( t ) ) = | t t | , t , t [ 0 , l ] .
In particular, g is an isometry and d ( p 1 , p 2 ) = l . The image α of g is said to be a geodesic segment (or metric segment) joining p 1 and p 2 . When it is unique, this geodesic segment is denoted by [ p 1 , p 2 ] . The space ( X , d ) is called a geodesic space if every two points of X are joined by a geodesic segment, and X is called a uniquely geodesic segment if there is exactly one geodesic segment joining p 1 and p 2 for each p 1 , p 2 X . A subset Y X is called convex if Y includes every geodesic segment joining any two of its points.
A geodesic triangle ( p 1 , p 2 , p 3 ) is a geodesic metric space ( X , d ) that consists of three vertices of Δ (the points p 1 , p 2 , p 3 X ) and the edges of Δ (a geodesic segment between each pair of vertices). A comparison triangle for the geodesic triangle ( p 1 , p 2 , p 3 ) in ( X , d ) is a triangle ¯ ( p 1 , p 2 , p 3 ) = ( p 1 ¯ , p 2 ¯ , p 3 ¯ ) in R 2 such that
d R 2 ( p i ¯ , p j ¯ ) = d ( p i , p j ) , i , j { 1 , 2 , 3 } .
A comparison triangle for the geodesic triangle always exists (see, [1,2]).
A geodesic metric space is called a C A T ( 0 ) space (this term is due to M. Gromov [3] and it is an acronym for E. Cartan, A.D. Aleksandrov and V.A. Toponogov) if all geodesic triangles of appropriate size satisfy the following C A T ( 0 ) comparison axiom.
Let Δ be a geodesic triangle in ( X , d ) and let ¯ R 2 be a comparison triangle for Δ. Then Δ is said to satisfy the C A T ( 0 ) inequality if for all vertices p 1 , p 2 and all comparison points p 1 ¯ , p 2 ¯ ¯ ,
d ( p 1 , p 2 ) d R 2 ( p 1 ¯ , p 2 ¯ ) .
Let p , p 1 , p 2 be points of a C A T ( 0 ) space, if p 0 is the midpoint of the segment [ p 1 , p 2 ] , which we will denote by p 1 p 2 2 , then the C A T ( 0 ) inequality implies
d 2 p , p 1 p 2 2 = d 2 ( p , p 0 ) 1 2 d 2 ( p , p 1 ) + 1 2 d 2 ( p , p 2 ) 1 4 d 2 ( p 1 , p 2 ) .
This inequality is called the (CN) inequality ([4]).
Remark 1.
A geodesic metric space ( X , d ) is a C A T ( 0 ) space if and only if it satisfies the (CN) inequality (cf. [1], p. 163).
The above (CN) inequality has been extended as
d 2 ( p , α p 1 ( 1 α ) p 2 ) α d 2 ( p , p 1 ) + ( 1 α ) d 2 ( p , p 2 ) α ( 1 α ) d 2 ( p 1 , p 2 ) , p , p 1 , p 2 X , ( CN * )
for all 0 α 1 [5,6].
In recent years, C A T ( 0 ) spaces have attracted many researchers as they treated a very important role in different directions of geometry and mathematics (see [1,7,8,9,10]). Complete C A T ( 0 ) spaces are often called Hadamard spaces (see [10]).
It is well known that a normed linear space satisfies the (CN) inequality if and only if it satisfies the parallelogram identity, i.e., it is a pre-Hilbert space [1]. Hence it is not so unusual to have an inner product-like notion in Hadamard spaces. In [11], they introduced the concept of quasilinearization as follows
Let us usually denote a pair ( p , q ) X 2 = X × X by p q and call it a vector. Then quasilinearization is defined as a mapping · , · : X 2 × X 2 R by
x y , u v = 1 2 ( d 2 ( x , v ) + d 2 ( y , u ) d 2 ( x , u ) d 2 ( y , v ) ) , x , y , u , v X .
It is easily seen that
x y , u v = u v , x y , x y , u v = y x , u v
and
x y , u v = x w , u v + w y , u v
for all x , y , u , v , w X . We say that X satisfies the Cauchy–Schwarz inequality if
x y , u v d ( x , y ) d ( u , v ) , x , y , u , v X .
Remark 2.
A geodesically connected metric space is a C A T ( 0 ) space if and only if it satisfies the Cauchy–Schwarz inequality ([11], Corollary 3).
In [12], the authors introduced the concept of duality mapping in C A T ( 0 ) spaces, by using the concept of quasilinearization, and studied its relation with the subdifferential. Moreover, they proved a characterization of metric projection in C A T ( 0 ) spaces as follows.
Theorem 1.
([12], Theorem 2.4) Let C be a nonempty convex subset of a complete C A T ( 0 ) space X . Then
p = P C x y p , p x 0 , y C ,
for all x X and p C .
In 2015, using the concept of quasilinearization, Wangkeeree et al. [13] proved the strong convergence theorems of the following Moudafi’s viscosity iterations for an asymptotically nonexpansive mapping T: For given a contraction mapping f defined on C and 0 < α n < 1 , let x n C be the unique fixed point of the contraction x α n f ( x ) ( 1 α n ) T n x , i.e.,
x n = α n f ( x n ) ( 1 α n ) T n x n , n 1
and let x 1 C be arbitrarily chosen and
x n + 1 = α n f ( x n ) ( 1 α n ) T n x n , n 1 .
They proved the iterative schemes { x n } defined by Equations (2) and (3) strongly converge to the same point x ¯ F ( T ) with x ¯ = P F ( T ) f ( x ¯ ) , which is the unique solution of the variational inequality
x ¯ f ( x ¯ ) , x x ¯ 0 , x F ( T ) ,
where F ( T ) = { x : T x = x } .
On the other hand, Shi et al. [14] studied the Δ-convergence of the iteration scheme for asymptotically nonexpansive mappings in C A T ( 0 ) spaces.
Let ( X , d ) be a metric space and C be a nonempty subset of X. A mapping f defined on C is called a contraction with coefficient 0 < α < 1 if
d ( f ( u ) , f ( v ) ) α d ( u , v )
for all u , v C . A subset C is called a retract of X if there exists a continuous mapping P from X onto C such that P u = u for all u C . A mapping P : X C is said to be a retraction if P 2 = P . Moreover, if a mapping P is a retraction, then P v = v for all v in the range of P.
Definition 1.
Let C be a nonempty subset of a metric space ( X , d ) . Let P : X C be a nonexpansive retraction of X onto C.
(1) 
A nonself mapping T : C X is said to be nonexpansive (cf. [15]) if
d ( T x , T y ) d ( x , y ) ,
for all x , y C .
(2) 
A nonself mapping T : C X is said to be asymptotically nonexpansive ([16]) if there exists a sequence { k n } [ 1 , ) with lim n k n = 1 such that
d ( T ( P T ) n 1 x , T ( P T ) n 1 y ) k n d ( x , y ) , n N ,
for all x , y C .
Recently, Kim et al. [17] and Kim [18] presented the existence and Δ-convergence for asymptotically nonexpansive nonself mappings in C A T ( 0 ) spaces.
Motivated and inspired by Wangkeeree et al. [13], Shi et al. [14], Kim et al. [17] and Kim [18], the aim of this paper is to obtain the strong convergence theorems of the Moudafi’s viscosity approximation methods for an asymptotically nonexpansive nonself mapping in C A T ( 0 ) spaces.
Let C be a nonempty closed convex subset of a complete C A T ( 0 ) space X. Let P : X C be a retraction mapping and T : C X be an asymptotically nonexpansive nonself mapping. Given a contraction mapping f defined on C and 0 < α n < 1 , let x n C be the unique fixed point of the contraction x α n f ( x ) ( 1 α n ) T ( P T ) n 1 x , i.e.,
x n = α n f ( x n ) ( 1 α n ) T ( P T ) n 1 x n , n 1
and let x 1 C be arbitrarily chosen and
x n + 1 = α n f ( x n ) ( 1 α n ) T ( P T ) n 1 x n , n 1 .
The author proved that the iterative schemes { x n } defined by Equations (4) and (5) strongly converge to the same point x * F ( T ) such that x * = P F ( T ) f ( x * ) which is the unique solution of the variational inequality
x * f ( x * ) , x x * 0 , x F ( T ) ,
where F ( T ) = { x : T x = x } .

2. Preliminaries

Throughout this paper, N denotes the set of all positive integers. Let C be a nonempty subset of a metric space ( X , d ) . F ( T ) = { x : T x = x } denotes the set of fixed points of T.
We write ( 1 t ) p 1 t p 2 for the unique point p in the geodesic segment joining from p 1 to p 2 such that
d ( p , p 1 ) = t d ( p 1 , p 2 ) and d ( p , p 2 ) = ( 1 t ) d ( p 1 , p 2 ) .
We also denote by [ p 1 , p 2 ] the geodesic segment joining from p 1 to p 2 , i.e., [ p 1 , p 2 ] = { ( 1 t ) p 1 t p 2 : T [ 0 , 1 ] } . A subset C of a C A T ( 0 ) space is convex if [ p 1 , p 2 ] C for all p 1 , p 2 C .
In the sequel we need the following useful lemmas.
Lemma 1.
([1], Proposition 2.2, p. 176) Let X be a C A T ( 0 ) space, then the distance function d : X × X R is convex, i.e., given any pair of geodesics g : [ 0 , 1 ] X and g : [ 0 , 1 ] X , parameterized proportional to arc length, the following inequality holds for all t [ 0 , 1 ] :
d ( g ( t ) , g ( t ) ) ( 1 t ) d ( g ( 0 ) , g ( 0 ) ) + t d ( g ( 1 ) , g ( 1 ) ) .
Lemma 2.
([6]) Let X be a C A T ( 0 ) space, p 1 , p 2 , z X and t [ 0 , 1 ] . Then
(i) 
d ( t p 1 ( 1 t ) p 2 , z ) t d ( p 1 , z ) + ( 1 t ) d ( p 2 , z ) ,
(ii) 
d 2 ( t p 1 ( 1 t ) p 2 , z ) t d 2 ( p 1 , z ) + ( 1 t ) d 2 ( p 2 , z ) t ( 1 t ) d 2 ( p 1 , p 2 ) .
Lemma 3.
([19]) Let X be a C A T ( 0 ) space, p 1 , p 2 , z X and t [ 0 , 1 ] . Then
(i) 
d ( t p 1 ( 1 t ) p 2 , γ p 1 ( 1 γ ) p 2 ) = | t γ | d ( p 1 , p 2 ) ,
(ii) 
d ( t p 1 ( 1 t ) p 2 , t p 1 ( 1 t ) z ) ( 1 t ) d ( p 2 , z ) .
Now, we give the concept of Δ-convergence and its some basic properties.
Kirk and Panyanak [20] insisted the concept of Δ-convergence in C A T ( 0 ) spaces that was introduced by Lim [21] in 1976 is very similar to the weak convergence in a Banach space setting.
Let x n be a bounded sequence in C A T ( 0 ) space X. For p X , we set
r ( p , x n ) = lim sup n d ( p , x n ) .
The asymptotic radius A r ( x n ) of x n is given by
A r ( x n ) = inf r ( p , x n ) : p X ,
and the asymptotic center A c ( x n ) of x n is the set
A c ( x n ) = p X : r ( p , x n ) = A r ( x n ) .
It is well known that asymptotic center A c ( x n ) consists of exactly one point (see, e.g., [22], Proposition 7, p. 767) in a complete C A T ( 0 ) space.
Definition 2.
([20]) A sequence { x n } in a complete C A T ( 0 ) space X is said to Δ-converge to x X if x is the unique asymptotic center of { u n } for every subsequence { u n } of { x n } , i.e., A c ( { u n } ) = { x } . In this case one can write
x n x or lim n x n = x
and call x the Δ-limit of { x n } .
Remark 3.
In a C A T ( 0 ) space, strong convergence in the metric implies Δ-convergence (see, [23,24]).
For any bounded sequence { z n } in a C A T ( 0 ) space X , there exists x * X such that
φ ( x * ) = inf φ ( x ) : x X ,
where
φ ( x ) = lim sup n d ( z n , x ) , x X .
Lemma 4.
([20]) Every bounded sequence in a complete C A T ( 0 ) space always has a Δ-convergent subsequence.
Now, we shall give the existence of a fixed point for asymptotically nonexpansive nonself mapping T : C X in a complete C A T ( 0 ) space.
Lemma 5.
([18]) Let C be a nonempty closed convex subset of a complete C A T ( 0 ) space X and let T : C X be an asymptotically nonexpansive nonself mapping with a sequence { k n } [ 1 , ) with lim n k n = 1 . Then T has a unique fixed point in C. Moreover, the set F ( T ) is a closed and convex subset of X .
Before we state the next lemma, we need the following notation
{ z n } x * φ ( x * ) = inf φ ( x ) : x C ,
where C is a nonempty closed convex subset that contains the bounded sequence { z n } and φ ( x ) = lim sup n d ( z n , x ) .
Lemma 6.
Let X be a C A T ( 0 ) space and C be a nonempty closed convex subset of X . Let T : C X be an asymptotically nonexpansive nonself mapping with a sequence { k n } [ 1 , ) with lim n k n = 1 . If
lim n d ( z n , T z n ) = 0 and { z n } x * ,
then we have
T ( x * ) = x * .
Proof. 
Since lim n d ( z n , T z n ) = 0 , we have
φ ( x ) = lim sup n d ( T ( P T ) m 1 z n , x ) , m 1 .
Hence
φ ( T ( P T ) m 1 x ) = lim sup n d ( T ( P T ) m 1 z n , T ( P T ) m 1 x ) lim sup n k m d ( z n , x ) = k m lim sup n d ( z n , x ) = k m φ ( x ) , x C .
In particular, we have
lim m φ ( T ( P T ) m 1 x * ) = lim m k m φ ( x * ) = φ ( x * ) .
From Lemma 2-(ii),
d 2 z n , x * T ( P T ) m 1 x * 2 1 2 d 2 ( z n , x * ) + 1 2 d 2 ( z n , T ( P T ) m 1 x * ) 1 4 d 2 ( x * , T ( P T ) m 1 x * ) , m , n 1 .
Taking m as fixed and lim sup n on both sides, we have
φ 2 x * T ( P T ) m 1 x * 2 1 2 φ 2 ( x * ) + 1 2 φ 2 ( T ( P T ) m 1 x * ) 1 4 d 2 ( x * , T ( P T ) m 1 x * ) , m 1 .
From the definition of x * , we obtain
φ 2 ( x * ) 1 2 φ 2 ( x * ) + 1 2 φ 2 ( T ( P T ) m 1 x * ) 1 4 d 2 ( x * , T ( P T ) m 1 x * ) , m 1 ,
which implies
d 2 ( x * , T ( P T ) m 1 x * ) 2 φ 2 ( T ( P T ) m 1 x * ) 2 φ 2 ( x * ) .
Taking lim m on both sides, from Equation (6), we get
lim m d ( x * , T ( P T ) m 1 x * ) = 0 ,
that is
lim m T ( P T ) m 1 x * = x * .
Since T P is a continuous mapping, we obtain
x * = lim m T ( P T ) m 1 x * = lim m T P ( T ( P T ) m 2 x * ) = ( T P ) x * = T x * .
Lemma 7.
([23], Theorem 2.6) Let X be a complete C A T ( 0 ) space, { x n } be a sequence in X and x X . Then { x n } Δ-converges to x if and only if
lim sup n x x n , x y 0 , y X .
Lemma 8.
([25]) Let { a n } , { b n } , { c n } and { λ n } be nonnegative sequences such that
a n + 1 ( 1 λ n ) a n + λ n b n + c n , n 0 ,
with { λ n } [ 0 , 1 ] , n = 0 λ n = , lim n b n = 0 and n = 0 c n < . Then lim n a n = 0 .
The following two useful lemmas can be found in [13].
Lemma 9.
([13]) Let X be a complete C A T ( 0 ) space. Then the following inequality holds
d 2 ( p , r ) d 2 ( q , r ) + 2 p q , p r , p , q , r X .
Lemma 10.
([13]) Let X be a C A T ( 0 ) space. For any l ( 0 , 1 ) and x , y X , let
x l = l x ( 1 l ) y .
Then, for all u , v X ,
(i) 
x l u , x l v l x u , x l v + ( 1 l ) y u , x l v ,
(ii) 
x l u , x v l x u , x v + ( 1 l ) y u , x v and
x l u , y v l x u , y v + ( 1 l ) y u , y v .

3. Main Results

In this section, we study the convergence theorems of Moudafi’s viscosity approximation methods for asymptotically nonexpansive nonself mapping T : C X in a complete C A T ( 0 ) space.
Theorem 2.
Let C be a nonempty closed convex subset of a complete C A T ( 0 ) space X and let T : C X be an asymptotically nonexpansive nonself mapping with a sequence { k n } [ 1 , ) with lim n k n = 1 . Let f be a contraction mapping defined on C with coefficient α ( 0 , 1 ) . Let { a n } be a real valued sequence with 0 < a n < 1 . If it satisfies the following conditions
(i) 
k n 1 a n < 1 α < a n ( k n α ) , n N ,
(ii) 
a n 0 , k n 1 a n 0 and | a n a n 1 | a n 0 as n ,
then the following statements hold.
(1) 
There exists x n such that
x n = a n f ( x n ) ( 1 a n ) T ( P T ) n 1 x n , n N .
(2) 
The sequence { x n } converges strongly to x * as n such that
x * = P F ( T ) f ( x * ) ,
which is equivalent to the following variational inequality:
x * f ( x * ) , x x * 0 , x F ( T ) .
Proof. I.
For each integer n 1 , we shall define a mapping F n : C X by
F n ( x ) = a n f ( x ) ( 1 a n ) T ( P T ) n 1 x , x C .
First, we show that F n is a contraction mapping. For any x , y C , by Lemma 1
d ( F n ( x ) , F n ( y ) ) = d ( a n f ( x ) ( 1 a n ) T ( P T ) n 1 x , a n f ( y ) ( 1 a n ) T ( P T ) n 1 y ) a n d ( f ( x ) , f ( y ) ) + ( 1 a n ) d ( T ( P T ) n 1 x , T ( P T ) n 1 y ) a n α d ( x , y ) + ( 1 a n ) k n d ( x , y ) = ( ( 1 a n ) k n + α a n ) d ( x , y ) .
From the condition (i), we have ( 1 a n ) k n + α a n < 1 . So F n is a contraction mapping. Thus there exists a unique z n C such that
z n = F n ( z n ) ,
that is
z n = a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n .
II. Next, we show that { z n } is bounded. From Lemma 5, there exists v C which is a fixed point of T with
v = T v = ( T P ) v .
Taking T P mapping on the both sides, we have
v = ( T P ) v = ( T P ) ( T P ) v = T ( P T ) P v = T ( P T ) v .
Continuing this process, we obtain
v = T ( P T ) n 1 v , n N .
For any v F ( T ) , we have
d ( z n , v ) = d ( a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n , v ) a n [ d ( f ( z n ) , f ( v ) ) + d ( f ( v ) , v ) ] + ( 1 a n ) d ( T ( P T ) n 1 z n , v ) α a n d ( z n , v ) + a n d ( f ( v ) , v ) + ( 1 a n ) k n d ( z n , v ) = ( k n ( k n α ) a n ) d ( z n , v ) + a n d ( f ( v ) , v ) .
Then
d ( z n , v ) a n ( k n α ) a n ( k n 1 ) d ( f ( v ) , v ) = 1 k n α k n 1 a n d ( f ( v ) , v ) 1 k n α ( k n 1 ) d ( f ( v ) , v ) = 1 1 α d ( f ( v ) , v ) .
Hence { z n } is bounded. So { T z n } and { f ( z n ) } are bounded. For v F ( T ) and Equation (10),
d ( T ( P T ) n 1 z n , v ) = d ( T ( P T ) n 1 z n , T ( P T ) n 1 v ) k n d ( z n , v ) L · d ( z n , v ) ,
where L = sup n k n . It follows that the sequence { T ( P T ) n 1 z n } is bounded.
III. We shall claim that
lim n d ( z n , T z n ) = 0 .
III-1. From Equation (9) and Lemma 2-(i), we get
d ( z n , T ( P T ) n 1 z n ) = d ( a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n , T ( P T ) n 1 z n ) a n d ( f ( z n ) , T ( P T ) n 1 z n ) .
On the other hand, since
d ( f ( z n ) , T ( P T ) n 1 z n ) d ( f ( z n ) , z n ) + d ( z n , T ( P T ) n 1 z n ) = d ( f ( z n ) , a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n ) + d ( z n , T ( P T ) n 1 z n ) ( 1 a n ) d ( f ( z n ) , T ( P T ) n 1 z n ) + d ( z n , T ( P T ) n 1 z n ) ,
we obtain
a n d ( f ( z n ) , T ( P T ) n 1 z n ) d ( z n , T ( P T ) n 1 z n ) .
Since a n 0 , from Equations (11) and (12), we have
lim n a n d ( f ( z n ) , T ( P T ) n 1 z n ) = lim n d ( z n , T ( P T ) n 1 z n ) = 0 .
III-2. By condition lim n k n 1 a n = 0 , for any 0 < ε < 1 α , there exists a sufficiently large n 0 , and we have
k n 1 a n ε .
From Equation (9) and Lemma 3, we have
d ( z n , z n 1 ) = d ( a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n , a n 1 f ( z n 1 ) ( 1 a n 1 ) T ( P T ) n 1 z n 1 ) d ( a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n , a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n 1 ) + d ( a n f ( z n ) ( 1 a n ) T ( P T ) n 1 z n 1 , a n f ( z n 1 ) ( 1 a n ) T ( P T ) n 1 z n 1 ) + d ( a n f ( z n 1 ) ( 1 a n ) T ( P T ) n 1 z n 1 , a n 1 f ( z n 1 ) ( 1 a n 1 ) T ( P T ) n 1 z n 1 ) ( 1 a n ) d ( T ( P T ) n 1 z n , T ( P T ) n 1 z n 1 ) + a n d ( f ( z n ) , f ( z n 1 ) + | a n a n 1 | d ( f ( z n 1 ) , T ( P T ) n 1 z n 1 ) ( 1 a n ) k n d ( z n , z n 1 ) + a n α d ( z n , z n 1 ) + | a n a n 1 | M * ,
where M * = sup n 1 d ( f ( z n 1 ) , T ( P T ) n 1 z n 1 ) . This implies that
( 1 ( 1 a n ) k n α a n ) d ( z n , z n 1 ) | a n a n 1 | M * .
From condition (i), we know
1 ( 1 a n ) k n α a n = a n ( k n α ) ( k n 1 ) > 0
and from Equation (14), we have
a n ( k n α ) ( k n 1 ) a n ( k n α ) a n ε = ( k n α ε ) a n ( 1 α ε ) a n .
Thus
d ( z n , z n 1 ) | a n a n 1 | 1 ( 1 a n ) k n α a n M * 1 1 α ε · | a n a n 1 | a n M * 0 , as n .
III-3. Therefore, from Equations (13) and (15), we get
d ( z n , T z n ) d ( z n , T ( P T ) n 1 z n ) + d ( T ( P T ) n 1 z n , T ( P T ) n 1 z n 1 ) + d ( T ( P T ) n 1 z n 1 , T z n ) d ( z n , T ( P T ) n 1 z n ) + k n d ( z n , z n 1 ) + d ( T ( P T ) 1 1 ( P T ) n 1 z n 1 , T ( P T ) 1 1 z n ) d ( z n , T ( P T ) n 1 z n ) + k n d ( z n , z n 1 ) + k 1 d ( ( P T ) n 1 z n 1 , z n ) d ( z n , T ( P T ) n 1 z n ) + k n d ( z n , z n 1 ) + k 1 [ d ( T ( P T ) n 2 z n 1 , z n 1 ) + d ( z n 1 , z n ) ] 0 , as n .
IV. Finally, we will show that { z n } contains a subsequence converge strongly to x * such that
x * = P F ( T ) f ( x * ) ,
which is equivalent to the following variational inequality
x * f ( x * ) , x x * 0 , x F ( T ) .
IV-1. Since { z n } is bounded, there exists a subsequence { z n i } of { z n } which Δ-converges to x * . By Lemmas 4 and 6, we may assume that { z n i } Δ-converges to a point x * and x * F ( T ) . It follows from Lemma 10-(i) and Equations (1) and (10) that
d 2 ( z n i , x * ) = z n i x * , z n i x * a n i f ( z n i ) x * , z n i x * + ( 1 a n i ) T ( P T ) n i 1 z n i x * , z n i x * a n i f ( z n i ) x * , z n i x * + ( 1 a n i ) d ( T ( P T ) n i 1 z n i , x * ) d ( z n i , x * ) a n i f ( z n i ) x * , z n i x * + ( 1 a n i ) k n i d 2 ( z n i , x * ) .
Since
f ( z n i ) x * , z n i x * = f ( z n i ) f ( x * ) + f ( x * ) x * , z n i x * = f ( z n i ) f ( x * ) , z n i x * + f ( x * ) x * , z n i x * d ( f ( z n i ) , f ( x * ) ) d ( z n i , x * ) + f ( x * ) x * , z n i x * α d 2 ( z n i , x * ) + f ( x * ) x * , z n i x * ,
combining Equation (16), it follows that
d 2 ( z n i , x * ) a n i α d 2 ( z n i , x * ) + a n i f ( x * ) x * , z n i x * + ( 1 a n i ) k n i d 2 ( z n i , x * ) .
Hence
d 2 ( z n i , x * ) a n i a n i ( k n i α ) ( k n i 1 ) f ( x * ) x * , z n i x * 1 1 α f ( x * ) x * , z n i x * .
Since { z n i } Δ-converges to x * , by Lemma 7, we have
lim sup i f ( x * ) x * , z n i x * 0 .
It follows from Equation (17) that { z n i } converges strongly to x * .
IV-2. Next, we will show that x * solves the variational inequality of Equation (8). Applying Lemma 2-(ii), for any z F ( T ) ,
d 2 ( z n i , z ) = d 2 ( a n i f ( z n i ) ( 1 a n i ) T ( P T ) n i 1 z n i , z ) a n i d 2 ( f ( z n i ) , z ) + ( 1 a n i ) d 2 ( T ( P T ) n i 1 z n i , z ) a n i ( 1 a n i ) d 2 ( f ( z n i ) , T ( P T ) n i 1 z n i ) a n i d 2 ( f ( z n i ) , z ) + ( 1 a n i ) k n i 2 d 2 ( z n i , z ) a n i ( 1 a n i ) d 2 ( f ( z n i ) , T ( P T ) n i 1 z n i ) .
Thus, we have
a n i ( 1 a n i ) d 2 ( f ( z n i ) , T ( P T ) n i 1 z n i ) + a n i k n i 2 d 2 ( z n i , z ) a n i d 2 ( f ( z n i ) , z ) + ( k n i 2 1 ) d 2 ( z n i , z ) ,
so
( 1 a n i ) d 2 ( f ( z n i ) , T ( P T ) n i 1 z n i ) + k n i 2 d 2 ( z n i , z ) d 2 ( f ( z n i ) , z ) + k n i 2 1 a n i d 2 ( z n i , z ) d 2 ( f ( z n i ) , z ) + k n i 1 a n i M ¯ ,
where M ¯ = ( L + 1 ) d 2 ( z n i , z ) , L = sup i 1 k n i . Since z n i x * and by Equation (13), we have
T ( P T ) n i 1 z n i x * .
From the conditions k n 1 , a n 0 , k n 1 a n 0 , continuity of the metric d and Equation (19), we have Equation (18) as follows
d 2 ( f ( x * ) , x * ) + d 2 ( x * , z ) d 2 ( f ( x * ) , z ) .
Therefore
0 1 2 ( d 2 ( x * , x * ) + d 2 ( f ( x * ) , z ) d 2 ( x * , z ) d 2 ( f ( x * ) , x * ) ) = x * f ( x * ) , z x * , z F ( T ) ,
that is, x * solves Equation (8).
IV-3. Finally, we will show the uniqueness of the solution of the variational inequality of Equation (8). Assume there exists a subsequence { z n j } of { z n } which Δ-converges to ω by the same argument. We know that ω F ( T ) and solves the variational inequality of Equation (8), i.e.,
x * f ( x * ) , x * ω 0
and
ω f ( ω ) , ω x * 0 .
From Equations (20) and (21), we can obtain
0 x * f ( x * ) , x * ω ω f ( ω ) , x * ω = x * f ( ω ) , x * ω + f ( ω ) f ( x * ) , x * ω ω x * , x * ω x * f ( ω ) , x * ω = x * ω , x * ω f ( ω ) f ( x * ) , ω x * x * ω , x * ω d ( f ( ω ) , f ( x * ) ) d ( ω , x * ) d 2 ( x * , ω ) α d 2 ( ω , x * ) = ( 1 α ) d 2 ( x * , ω ) .
Since 0 < α < 1 , we have
d ( x * , ω ) = 0 ,
so
x * = ω .
Hence { z n } converges strongly to x * , which solves the variational inequality of Equation (8). □
Now, we explain a strong convergence theorem for an asymptotically nonexpansive nonself mapping.
Theorem 3.
Let C be a nonempty closed convex subset of a complete C A T ( 0 ) space X and let T : C X be an asymptotically nonexpansive nonself mapping with a sequence { k n } [ 1 , ) with lim n k n = 1 . Let f be a contraction mapping defined on C with coefficient α ( 0 , 1 ) . For the arbitrarily given initial point x 0 C , let { x n } be a sequence generated by
x n + 1 = a n f ( x n ) ( 1 a n ) T ( P T ) n 1 x n , n 0
where { a n } ( 0 , 1 ) satisfies the following conditions:
(i) 
lim n a n = 0 ,
(ii) 
lim n k n 1 a n = 0 .
Then the sequence { x n } converges strongly to x * as n such that
x * = P F ( T ) f ( x * ) ,
which is equivalent to the variational inequality of Equation (8).
Proof. I.
First, we show that the sequence { x n } is bounded. From Lemma 5, there exists q C which is a fixed point of T with
q = T q = ( T P ) q = T ( P T ) n 1 q .
Since lim n k n 1 a n , for any 0 < ε < 1 α , there exists a sufficiently large n 0 , we have
k n 1 a n ε .
For any q F ( T ) , from Equations (13) and (22), we get
d ( x n + 1 , q ) = d ( a n f ( x n ) ( 1 a n ) T ( P T ) n 1 x n , q ) a n d ( f ( x n ) , q ) + ( 1 a n ) d ( T ( P T ) n 1 x n , q ) a n ( d ( f ( x n ) , f ( q ) ) + d ( f ( q ) , q ) ) + ( 1 a n ) k n d ( x n , q ) a n α d ( x n , q ) + a n d ( f ( q ) , q ) + ( 1 a n ) k n d ( x n , q ) = ( 1 + ( k n 1 ) a n ( k n α ) ) d ( x n , q ) + a n d ( f ( q ) , q ) ( 1 a n ( k n α ε ) ) d ( x n , q ) + a n d ( f ( q ) , q ) ( 1 a n ( 1 α ε ) ) d ( x n , q ) + a n d ( f ( q ) , q ) max d ( x n , q ) , 1 1 α ε d ( f ( q ) , q ) ,
for 0 a n ( 1 α ε ) 1 . Similarly, we can get
d ( x n , q ) max d ( x n 1 , q ) , 1 1 α ε d ( f ( q ) , q ) .
Continuing this process, we obtain that
d ( x n + 1 , q ) max d ( x 0 , q ) , 1 1 α ε d ( f ( q ) , q ) , n 0 .
Thus, the sequence { x n } is bounded. So { f ( x n ) } and { T x n } are also bounded. From the fact that { x n } is bounded and from Lemmas 4 and 6, there exists a subsequence { x n i } of { x n } which Δ-converges to q F ( T ) .
II. Next, we prove that x n q as n . For any n N , we set
y n = a n q + ( 1 a n ) T ( P T ) n 1 x n .
It follows from Lemmas 9 and 10 that
d 2 ( x n + 1 , q ) d 2 ( y n , q ) + 2 x n + 1 y n , x n + 1 q ( a n d ( q , q ) + ( 1 a n ) d ( T ( P T ) n 1 x n , q ) ) 2 + 2 a n f ( x n ) y n , x n + 1 q + ( 1 a n ) T ( P T ) n 1 x n y n , x n + 1 q ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + 2 [ a n 2 f ( x n ) q , x n + 1 q + a n ( 1 a n ) f ( x n ) T ( P T ) n 1 x n , x n + 1 q + ( 1 a n ) a n T ( P T ) n 1 x n q , x n + 1 q + ( 1 a n ) 2 T ( P T ) n 1 x n T ( P T ) n 1 x n , x n + 1 q ] = ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + 2 [ a n 2 f ( x n ) q , x n + 1 q + a n ( 1 a n ) f ( x n ) T ( P T ) n 1 x n , x n + 1 q + ( 1 a n ) a n T ( P T ) n 1 x n q , x n + 1 q ] = ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + 2 a n 2 f ( x n ) q , x n + 1 q + a n ( 1 a n ) f ( x n ) q , x n + 1 q = ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + 2 a n f ( x n ) q , x n + 1 q = ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + 2 a n f ( x n ) f ( q ) , x n + 1 q + 2 a n f ( q ) q , x n + 1 q ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + 2 a n α d ( x n , q ) d ( x n + 1 , q ) + 2 a n d ( f ( q ) , q ) d ( x n + 1 , q ) ( 1 a n ) 2 k n 2 d 2 ( x n , q ) + a n α ( d 2 ( x n , q ) + d 2 ( x n + 1 , q ) ) + 2 a n d ( f ( q ) , q ) d ( x n + 1 , q )
which implies
( 1 α a n ) d 2 ( x n + 1 , q ) ( ( 1 a n ) 2 k n 2 + α a n ) d 2 ( x n , q ) + 2 a n d ( f ( q ) , q ) d ( x n + 1 , q ) ,
d 2 ( x n + 1 , q ) ( 1 a n ) 2 k n 2 + α a n 1 α a n d 2 ( x n , q ) + 2 a n 1 α a n d ( f ( q ) , q ) d ( x n + 1 , q ) ( 1 a n ) k n 2 + α 1 α a n d 2 ( x n , q ) + 2 a n 1 α a n d ( f ( q ) , q ) d ( x n + 1 , q ) = 1 a n ( k n 2 α ) + ( 1 k n 2 α ) 1 α a n d 2 ( x n , q ) + 2 a n 1 α a n d ( f ( q ) , q ) d ( x n + 1 , q ) .
Now, taking
λ n = a n ( k n 2 α ) + ( 1 k n 2 α ) 1 α a n , b n = 2 a n a n ( k n 2 α ) + ( 1 k n 2 α ) d ( f ( q ) , q ) d ( x n + 1 , q ) ,
by Lemma 8, we can conclude that
lim n x n = q .
III. Finally, from the proof of IV-2 and IV-3 in Theorem 2, we can easily show that q F ( T ) is the unique solution satisfying the variational inequality of Equation (8). This completes the proof of Theorem 3. □
If a mapping T : C C is a self mapping, then P becomes the identity mapping. Thus we have the following corollaries (cf. [13,26]).
Corollary 1.
Let C be a nonempty closed convex subset of a complete C A T ( 0 ) space X and let T : C C be an asymptotically nonexpansive mapping with a sequence { k n } [ 1 , ) with lim n k n = 1 . Let f be a contraction defined on C with coefficient 0 < α < 1 . Let { a n } be a sequence of real numbers with 0 < a n < 1 . If it satisfies the following conditions
(i) 
k n 1 a n < 1 α < a n ( k n α ) , n N ,
(ii) 
a n 0 , k n 1 a n 0 and | a n a n 1 | a n 0 as n ,
then the following statements hold.
(1) 
There exists x n such that
x n = a n f ( x n ) ( 1 a n ) T n x n , n N .
(2) 
The sequence { x n } converges strongly to x * as n such that
x * = P F ( T ) f ( x * ) ,
which is equivalent to the following variational inequality:
x * f ( x * ) , x x * 0 , x F ( T ) .
Corollary 2.
Let C be a nonempty closed convex subset of a complete C A T ( 0 ) space X and let T : C C be an asymptotically nonexpansive mapping with a sequence { k n } [ 1 , ) with lim n k n = 1 . Let f be a contraction defined on C with coefficient 0 < α < 1 . For the arbitrary initial point x 0 C , let { x n } be generated by
x n + 1 = a n f ( x n ) ( 1 a n ) T n x n , n 0
where { a n } ( 0 , 1 ) satisfies the following conditions:
(i) 
lim n a n = 0 ,
(ii) 
lim n k n 1 a n = 0 .
Then the sequence { x n } converges strongly to x * as n such that
x * = P F ( T ) f ( x * ) ,
which is equivalent to the variational inequality of Equation (8).

4. Conclusions

Theorems 2 and 3 generalize and improve the results which are discussed in Wangkeeree et al. [13], Shi et al. [14], Kim et al. [17], Kim [18] and others.
The strong convergence theorems of the Moudafi’s viscosity approximation methods apply various classes of variational inequalities and optimization problems, its results proved in this paper continue to hold for these problems. It is expected that this class will inspire and motivate further research in this area.

Funding

This research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education(2019R1I1A3A01060344).

Acknowledgments

The author would like to thank the referees for their valuable comments and suggestions which improved the presentation of this paper.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Bridson, M.; Haefliger, A. Metric Spaces of Non-Positive Curvature; Springer: Berlin/Heidelberg, Germany, 1999. [Google Scholar]
  2. Saluja, G.S.; Kim, J.K. On the convergence of modified S-iteration process for asymptotically quasi-nonexpansive type mappings in a CAT(0) space. Nonlinear Funct. Anal. Appl. 2014, 19, 329–339. [Google Scholar] [CrossRef]
  3. Gromov, M. Hyperbolic Groups, Essays in Group Theory; Springe, Mathematical Sciences Research Institute Publications: New York, NY, USA, 1987; Volume 8. [Google Scholar]
  4. Bruhat, F.; Tits, J. Groups réductifss sur un corps local. I. Données radicielles valuées. Publ. Math. Inst. Hautes Eétudes Sci. 1972, 41, 5–251. [Google Scholar] [CrossRef]
  5. Khamsi, M.A.; Kirk, W.A. On uniformly Lipschitzian multivalued mappings in Banach and metric spaces. Nonlinear Anal. 2010, 72, 2080–2085. [Google Scholar] [CrossRef]
  6. Dhompongsa, S.; Panyanak, B. On Δ-convergence theorems in CAT(0) spaces. Comput. Math. Anal. 2008, 56, 2572–2579. [Google Scholar]
  7. Brown, K.S. Building; Springer: New York, NY, USA, 1989. [Google Scholar]
  8. Burago, D.; Burago, Y.; Ivanov, S. A course in metric Geometry. In Graduate studies in Mathematics; American Mathematical Society: Providence, RI, USA, 2001; Volume 33. [Google Scholar]
  9. Goebel, K.; Reich, S. Uniform Convexity, Hyperbolic Geometry, and Nonexpansive Mappings; Series of Monographs and Textbooks in Pure and Applied Mathematics; Dekker: New York, NY, USA, 1984; Volume 83. [Google Scholar]
  10. Kirk, W.A. A fixed point theorem in CAT(0) spaces and R-trees. Fixed Point Theory Appl. 2004, 2004, 309–316. [Google Scholar] [CrossRef]
  11. Berg, I.D.; Nikoleav, I.G. Quasilinearization and curvature of Alexandrov spaces. Geom. Dedic. 2008, 133, 195–218. [Google Scholar] [CrossRef]
  12. Dehghan, H.; Rooin, J. A characterization of metric projection in CAT(0) spaces. In Proceedings of the International Conference on Functional Equation, Geometric Functions and Applications (ICFGA 2012), Payame Noor University, Tabriz, Iran, 10–12 May 2012; pp. 41–43. [Google Scholar]
  13. Wangkeeree, R.; Boonkong, U.; Preechasilp, P. Viscosity approximation methods for asymptotically nonexpansive mapping in CAT(0) spaces. Fixed Point Theory Appl. 2015, 2015. [Google Scholar] [CrossRef]
  14. Shi, L.Y.; Chen, R.D.; Wu, Y.J. Δ-convergence problems for asymptotically nonexpansive mappings in CAT(0) spaces. Abstr. Appl. Anal. 2013, 2013, 251705. [Google Scholar] [CrossRef] [PubMed]
  15. Kim, K.S. Approximation of common fixed points for an implicit iteration with errors of finite family of nonself nonexpansive mappings in Banach spaces. Dyn. Contin. Discret. Impulse Syst. Ser. B (Appl. Algorithms) 2008, 15, 807–815. [Google Scholar]
  16. Chidume, C.E.; Ofoedu, E.U.; Zegeye, H. Strong and weak convergence theorems for asymptotically nonexpansive mappings. J. Math. Anal. Appl. 2003, 280, 364–374. [Google Scholar] [CrossRef]
  17. Kim, J.K.; Pathak, R.P.; Dashputre, S.; Diwan, S.D.; Gupta, R. Demiclosedness principle and convergence theorems for Lipshitzian type nonself-mappings in CAT(0) spaces. Nonlinear Funct. Anal. Appl. 2018, 23, 73–95. [Google Scholar]
  18. Kim, K.S. Existence theorem for asymptotically nonexpansive nonself mapping in CAT(0) spaces. Nonlinear Funct. Anal. Appl. 2019. submitted. [Google Scholar]
  19. Chaoha, P.; Phon-on, A. A note on fixed point sets in CAT(0) spaces. J. Math. Anal. Appl. 2006, 320, 983–987. [Google Scholar] [CrossRef]
  20. Kirk, W.A.; Panyanak, B. A concept of convergence in geodesic spaces. Nonlinear Anal. 2008, 68, 3689–3696. [Google Scholar] [CrossRef]
  21. Lim, T.C. Remarks on some fixed point theorems. Proc. Am. Math. Soc. 1976, 60, 179–182. [Google Scholar] [CrossRef]
  22. Dhompongsa, S.; Kirk, W.A.; Sims, B. Fixed point of uniformly Lipschitzian mappings. Nonlinear Anal. 2006, 65, 762–772. [Google Scholar] [CrossRef]
  23. Kakavandi, B.A. Weak topologies in complete CAT(0) metric spaces. Proc. Am. Math. Soc. 2013, 141, 1029–1039. [Google Scholar] [CrossRef]
  24. Kim, K.S. Equivalence between some iterations in CAT(0) spaces. J. Comput. Anal. Appl. 2018, 24, 474–485. [Google Scholar]
  25. Liu, L.S. Ishikawa and Mann iterative process with errors for nonlinear strongly accretive mappings in Banach spaces. J. Math. Anal. Appl. 1995, 194, 114–125. [Google Scholar] [CrossRef]
  26. Wangkeeree, R.; Preechasilp, P. Viscosity approximation methods for nonexpansive mappings in CAT(0) spaces. J. Inequal. Appl. 2013, 2013. [Google Scholar] [CrossRef]
Back to TopTop