ATTRACTIVE AND MEAN CONVERGENCE THEOREMS FOR TWO COMMUTATIVE NONLINEAR MAPPINGS IN BANACH SPACES.

1. INTRODUCTION

Let H be a real Hilbert space and let C be a nonempty subset of H. Let T be a mapping of C into H. Then we denote by F(T) the set of fixed points of T and by A(T) the set of attractive points [27] of T, i.e.,

(i) F(T) = {z [member of] C: Tz = z};

(ii) A(T) = {z [member of] H : [parallel]Tx - z[parallel] [less than or equal to] [parallel]x - z[parallel], [for all]x [member of] C}.

We know from [27] that A(T) is closed and convex. This property is important for proving mean convergence theorems. Such a concept of attractive points was defined in a Banach space; see [20]. A mapping T : C [right arrow] H is said to be nonexpansive [4] if [parallel]Tx - Ty[parallel] [less than or equal to] [parallel]x - y[parallel] for all x,y [member of] C. Baillon [2] proved the first mean convergence [parallel]Tx - Ty[parallel] [less than or equal to] [parallel]x - y[parallel] for all x,y [member of] C. Baillon [2] proved the first mean convergence theorem in a Hilbert space.

Theorem 1.1 ([2]). Let C be a bounded, closed and convex subset of H and let T:C [right arrow] C be nonexpansive. Then for any x [member of] C,

[S.sub.n]x = 1/n [n-1.summation over (k=0)] [T.sup.k] x

converges weakly to an element z [member of] F(T).

This theorem for nonexpansive mappings has been extended to Banach spaces by many authors; see, for example, [3, 5]. On the other hand, in 2010, Kocourek, Takahashi and Yao [13] defined a broad class of nonlinear mappings in a Hilbert space: Let H be a Hilbert space and let C be a nonempty subset of H. A mapping T : C [right arrow] H is called generalized hybrid [13] if there exist [alpha], [beta] [member of] R such that

(1.1) [mathematical expression not reproducible]

for all x,y [member of] C. Such a mapping T is called ([alpha], [beta])- generalized hybrid. Notice that the class of generalized hybrid mappings covers several well-known mappings. For example, a (1,0)-generalized hybrid mapping is nonexpansive, i.e.,

[parallel]Tx - Ty[parallel] [less than or equal to] [parallel]x - y[parallel], [for all] x,y [member of] C.

It is nonspreading [17, 18] for [alpha] = 2 and [beta] = 1, i.e.,

[mathematical expression not reproducible].

It is also hybrid [25] for [alpha] = 3/2 and [beta] = 1/2, i.e.,

[mathematical expression not reproducible].

In general, nonspreading and hybrid mappings are not continuous; see [10]. The mean convergence theorem by Baillon [2] for nonexpansive mappings has been extended to generalized hybrid mappings in a Hilbert space by Kocourek, Takahashi and Yao [13]. Furthermore, Takahashi and Takeuchi [27] proved the following mean convergence theorem without convexity in a Hilbert space.

Theorem 1.2 ([27]). Let H be a Hilbert space and let C be a nonempty subset of H. Let T be a generalized hybrid mapping from C into itself. Assume that {[T.sup.n]z} for some z [member of] C is bounded and define

[S.sub.n]x = 1/n [n-1.summation over (k=0)] [T.sup.k]x

for all x [member of] C and n [member of] N. Then {[S.sub.n]x} converges weakly to [u.sub.0] G A(T), where [u.sub.0] = [lim.sub.n[right arrow][infinity]] [P.sub.A(T)][T.sup.n]x and [P.sub.A(T)] is the metric projection of H onto A(T).

Maruyama, Takahashi and Yao [21] also defined a more broad class of nonlinear mappings called 2-generalized hybrid which covers generalized hybrid mappings in a Hilbert space. Let C be a nonempty subset of H and let T be a mapping of C into H. A mapping T:C [right arrow] H is 2-generalized hybrid [21] if there exist [[alpha].sub.1], [[alpha].sub.2], [[beta].sub.1], [[beta].sub.2] [member of] R such that

(1.2) [mathematical expression not reproducible]

for all x, y [member of] C.

Recently, Hojo, Takahashi and Takahashi [6] proved attractive and mean convergence theorems without convexity for commutative 2-generalized hybrid mappings in a Hilbert space. These results generalize Takahashi and Takeuchi's theorem (Theorem 1.2) and Kohsaka's theorem [15] which is a mean convergence theorem for commutative [lambda]-hybrid mappings in a Hilbert space.

In this paper, using the class of 2-generalized nonspreading mappings which was defined by [29] in a Banach space and covers 2-generalized hybrid mappings in a Hilbert space, we prove an attractive point theorem in a Banach space. This theorem generalizes Hojo, Takahashi and Takahashi's attractive point theorem [6] in a Hilbert space. Then we prove a mean convergence theorem of Baillon's type [2] without convexity for commutative 2-generalized nonspreading mappings in a Banach space. This result is a general mean convergence theorem which extends Baillon's theorem (Theorem 1.1) to a Banach space.

2. PRELIMINARIES

Let E be a real Banach space with norm [parallel]*[parallel] and let [E.sub.*] be the topological dual space of E. We denote the value of [y.sup.*] [member of] [E.sup.*] at x [member of] E by <x, [y.sup.*]>. When {[x.sub.n]} is a sequence in E, we denote the strong convergence of {[x.sub.n]} to x G [member of] by [x.sub.n] [right arrow] x and the weak convergence by [x.sub.n] [??] x. The modulus [delta] of convexity of E is defined by

[mathematical expression not reproducible]

for every [epsilon] with 0 [less than or equal to] [epsilon] [less than or equal to] 2. A Banach space E is said to be uniformly convex if [delta]([epsilon]) > 0 for every [epsilon] > 0. A uniformly convex Banach space is strictly convex and reflexive. Let C be a nonempty subset of a Banach space E. A mapping T : C [right arrow] E is nonexpansive if [parallel]Tx - Ty[parallel] [less than or equal to] [parallel]x - y[parallel] for all x, y [member of] C. A mapping T : C [right arrow] E is quasi-nonexpansive if F(T) = 0 and [parallel]Tx - y[parallel] [less than or equal to] [parallel]x - y[parallel] for all x [member of] C and y [member of] F(T), where F(T) is the set of fixed points of T. If C is a nonempty, closed and convex subset of a strictly convex Banach space E and T : C [right arrow] E is quasi- nonexpansive, then F(T) is closed and convex; see Itoh and Takahashi [11].

Let E be a Banach space. The duality mapping J from E into [mathematical expression not reproducible] is defined by

Jx = {[x.sup.*] [member of] [E.sup.*]: <x,[x.sup.*]> = [[parallel]x[parallel].sup.2] = [[parallel][x.sup.*][parallel].sup.2]}

for every x [member of] E. Let U = {x [member of] E: [parallel]x[parallel] = 1}. The norm of E is said to be Gateaux differentiable if for each x, y [member of] U, the limit

(2.1) [mathematical expression not reproducible]

exists. In this case, E is called smooth. We know that E is smooth if and only if J is a single-valued mapping of E into [E.sup.*]. We also know that E is reflexive if and only if J is surjective, and E is strictly convex if and only if J is one-to-one. Therefore, if E is a smooth, strictly convex and reflexive Banach space, then J is a single- valued bijection. The norm of E is said to be uniformly Gateaux differentiable if for each y [member of] U, the limit (2.1) is attained uniformly for x E U. It is also said to be Frechet differentiable if for each x [member of] U, the limit (2.1) is attained uniformly for y [member of] U. A Banach space E is called uniformly smooth if the limit (2.1) is attained uniformly for x,y [member of] U. It is known that if the norm of E is uniformly Gateaux differentiable, then J is uniformly norm to weak * continuous on each bounded subset of E, and if the norm of E is Frechet differentiable, then J is norm to norm continuous. If E is uniformly smooth, J is uniformly norm to norm continuous on each bounded subset of E. For more details, see [23, 24].

Lemma 2.1 ([23, 24]). Let E be a smooth Banach space and let J be the duality mapping on E. Then <x - y, Jx - Jy> [greater than or equal to] 0 for all x,y [member of] E. Furthermore, if E is strictly convex and <x - y, Jx - Jy> = 0, then x = y.

Let E be a smooth Banach space. The function [phi]: E x E [right arrow](- [infinity], [infinity]) is defined by

(2.2) [phi](x,y) = [[parallel]x[parallel].sup.2] - 2<x, Jy> + [[parallel]y[parallel].sup.2]

for x, y [member of] E, where J is the duality mapping of E; see [1] and [12]. We have from the definition of f that

(2.3) [phi](x y) = [phi](x z) + [phi](z, y) + 2<x - z, Jz - Jy>

for all x, y, z [member of] E. From [([parallel]x[parallel] - [parallel]y[parallel]).sup.2] [less than or equal to] [phi](x,y) for all x,y [member of] E, we can see that [phi] (x, y) [greater than or equal to] 0. Furthermore, we can obtain the following equality:

(2.4) 2<x - y, Jz - Jw> = [phi](x, w) + [phi](y, z) - [phi](x, z) - [phi](y, w)

for x, y, z, w [member of] E. If E is additionally assumed to be strictly convex, then from

Lemma 2.1 we have

(2.5) [phi] (x, y) = 0 [??] x = y.

The following lemma is in Xu [33].

Lemma 2.2 ([33]). Let E be a uniformly convex Banach space and let r > 0. Then there exists a strictly increasing, continuous and convex function g : [0, [infinity]) [right arrow] [0, [infinity]) such that g(0) = 0 and

[mathematical expression not reproducible]

for all x, y [member of] [B.sub.r] and [lambda] with 0 [less than or equal to] [lambda] [less than or equal to] 1, where [B.sub.r] = {z [member of] E : [parallel]z[parallel] [less than or equal to] r}.

Using Lemma 2.2, we have the following lemma by Kamimura and Takahashi [12].

Lemma 2.3 ([12]). Let E be a smooth and uniformly convex Banach space and let r > 0. Then there exists a strictly increasing, continuous and convex function g : [0, 2r] [right arrow] R such that g(0) = 0 and

g([parallel]x - y[parallel]) [less than or equal to] [phi](x,y)

for all x, y [member of] Br, where [B.sub.r] = {z [member of] E : [parallel]z[parallel] [less than or equal to] r}.

Let E be a smooth Banach space. Let C be a nonempty subset of E and let T be a mapping of C into E. We denote by A(T) the set of attractive points of T, i.e., A(T) = {z [member of] E: [phi](z, Tx) [less than or equal to] [phi](z, x), [for all]x [member of] C}; see [20].

Lemma 2.4 ([20]). Let E be a smooth Banach space and let C be a nonempty subset of E. Let T be a mapping from C into E. Then A(T) is a closed and convex subset of E.

Let E be a smooth Banach space and let C be a nonempty subset of E. Then a mapping T : C [right arrow] E is called generalized nonexpansive [8] if F(T) [not equal to] 0 and

[phi](Tx,y) [less than or equal to] [phi](x, y)

for all x [member of] C and y [member of] F(T); see also [32]. Let D be a nonempty subset of a Banach space E. A mapping R : E [right arrow] D is said to be sunny if

R(Rx + t(x - Rx)) = Rx

for all x [member of] E and t [greater than or equal to] 0. A mapping R : E [right arrow] D is said to be a retraction or a projection if Rx = x for all x [member of] D. A nonempty subset D of a smooth Banach space E is said to be a generalized nonexpansive retract (resp. sunny generalized nonexpansive retract) of E if there exists a generalized nonexpansive retraction (resp. sunny generalized nonexpansive retraction) R from E onto D; see [8] for more details. The following results are in Ibaraki and Takahashi [8].

Lemma 2.5 ([8]). Let C be a nonempty closed sunny generalized nonexpansive retract of a smooth and strictly convex Banach space E. Then the sunny generalized nonexpansive retraction from E onto C is uniquely determined.

Lemma 2.6 ([8]). Let C be a nonempty closed subset of a smooth and strictly convex Banach space E such that there exists a sunny generalized nonexpansive retraction R from E onto C and let (x, z) [member of] E x C. Then the following hold:

(i) z = Rx if and only if <x - z, Jy - Jz> [less than or equal to] 0 for all y [member of] C;

(ii) [phi](Rx, z) + [phi](x, Rx) [less than or equal to] [phi](x, z).

In 2007, Kohsaka and Takahashi [16] proved the following results:

Lemma 2.7 ([16]). Let E be a smooth, strictly convex and reflexive Banach space and let C be a nonempty closed subset of E. Then the following are equivalent:

(a) C is a sunny generalized nonexpansive retract of E;

(b) C is a generalized nonexpansive retract of E;

(c) JC is closed and convex.

Lemma 2.8 ([16]). Let E be a smooth, strictly convex and reflexive Banach space and let C be a nonempty closed sunny generalized nonexpansive retract of E. Let R be the sunny generalized nonexpansive retraction from E onto C and let (x, z) [member of] E x C. Then the following are equivalent:

(i) z = Rx;

(ii) [phi](x,z) = [min.sub.y [member of] C] f(x, y).

Ibaraki and Takahashi [9] also obtained the following result concerning the set of fixed points of a generalized nonexpansive mapping.

Lemma 2.9 ([9]). Let E be a reflexive, strictly convex and smooth Banach space and let T be a generalized nonexpansive mapping from E into itself. Then F(T) is closed and JF(T) is closed and convex.

The following theorem is proved by using Lemmas 2.7 and 2.9.

Lemma 2.10 ([9]). Let E be a reflexive, strictly convex and smooth Banach space and let T be a generalized nonexpansive mapping from E into itself. Then F(T) is a sunny generalized nonexpansive retract of E.

Using Lemma 2.7, we also have the following result.

Lemma 2.11 ([26]). Let E be a smooth, strictly convex and reflexive Banach space and let {[C.sub.i]: i [member of] I} be a family of sunny generalized nonexpansive retracts of E such that [[intersection].sub.i [member of] I][C.sub.i] is nonempty. Then [[intersection].sub.i [member of] I] [C.sub.i] is a sunny generalized nonexpansive retract of E.

Let [l.sup.[infinity]] be the Banach space of bounded sequences with supremum norm. Let [mu] be an element of [([l.sup.[infinity]]).sup.*] (the dual space of [l.sup.[infinity]]). Then, we denote by [mu](f) the value of [mu] at f = ([x.sub.1], [x.sub.2], [x.sub.3], ...) G [l.sup.[infinity]]. Sometimes, we denote by [[mu].sub.n]([x.sub.n]) the value [mu](f). A linear functional [mu] on [l.sup.[infinity]] is called a mean if [mu](e) = [parallel]a[parallel] = 1, where e = (1,1,1, ...). A mean [mu] is called a Banach limit on [l.sup.[infinity]] if [[mu].sub.n]([x.sub.n+1]) = [[mu].sub.n]([x.sub.n]). We know that there exists a Banach limit on [l.sup.[infinity]]. If [mu] is a Banach limit on [l.sup.[infinity]], then for f = ([x.sub.1], [x.sub.2], [x.sub.3], ...) [member of] [l.sup.[infinity]],

[mathematical expression not reproducible].

In particular, if f = ([x.sub.1], [x.sub.2], [x.sub.3], ...) [member of] [l.sup.[infinity]] and [x.sub.n] [right arrow] a [member of] R, then we have [mu](f) = [[mu].sub.n]([x.sub.n]) = a. For the proof of existence of a Banach limit and its other elementary properties, see [23].

3. FIXED POINT THEOREMS

Let E be a smooth Banach space and let C be a nonempty subset of E. Then a mapping T : C [right arrow] E is called 2-generalized nonspreading [29] if there exist [[alpha].sub.1],[[alpha].sub.2], [[beta].sub.1], [[beta].sub.2], [[gamma].sub.1], [[gamma].sub.2], [[delta].sub.1], [[delta].sub.2] [member of] R such that

(3.1) [mathematical expression not reproducible]

for all x,y [member of] C; see also [30]. Such a mapping is called ([[alpha].sub.1], [[alpha].sub.2], [[beta].sub.1], [[beta].sub.1], [[gamma].sub.1], [[gamma].sub.2], [[delta].sub.1], [[delta].sub.1])-generalized nonspreading. We know that a (0, [[alpha].sub.2], 0, [[beta].sub.2], 0, [[gamma].sub.2], 0, [[delta].sub.2])- generalized nonspreading mapping is generalized nonspreading in the sense of [14]. We also know that a (0,1, 0,1, 0,1, 0, 0)-generalized nonspreading mapping is nonspreading in the sense of [18].

Now we prove an attractive point theorem for commutative 2-generalized non- spreading mappings in a Banach space. Before proving it, we prove the following result.

Lemma 3.1. Let E be a smooth, strictly convex and reflexive Banach space with the duality mapping J and let C be a nonempty subset of E. Let S and T be mappings of C into itself. Let {[x.sub.n]} be a bounded sequence of E and let [mu] be a mean on [l.sup.[infinity]]. Suppose that

[mathematical expression not reproducible]

for all y [member of] C. Then A(S) [intersection] A(T) is nonemppty. Additionally, if C is closed and convex and {[x.sub.n]} [subset] C, then F(S) [intersection] F(T) is nonempty.

Proof. Using a mean [mu] and a bounded sequence {[x.sub.n]}, we define a function g : [E.sup.*] [right arrow] R as follows:

g([x.sup.*]) = [[mu].sub.n]<[x.sub.n], [x.sup.*]>

for all [x.sup.*] [member of] [E.sup.*]. Since [mu] is linear, g is also linear. Furthermore, we have

[mathematical expression not reproducible]

for all [x.sup.*] [member of] [E.sup.*]. Then g is a linear and bounded real- valued function on [E.sup.*]. Since E is reflexive, there exists a unique element z of E such that

g([x.sup.*]) = [[mu].sub.n] <[x.sub.n], [x.sup.*]> = <z, [x.sup.*]>

for all [x.sup.*] [member of] [E.sup.*]. From (2.3) we have that for y [member of] C and n [member of] N,

[phi]([x.sub.n], y) = [phi]([x.sub.n], Sy) + [phi](Sy, y) + 2<[x.sub.n] - Sy, JSy - Jy>.

So, we have that for y [member of] C,

[mathematical expression not reproducible].

Since, by assumption, [[mu].sub.n][phi]([x.sub.n], Sy) [less than or equal to] [[mu].sub.n][phi]([x.sub.n], y) for all y [member of] C, we have

[[mu].sub.n] [phi]([x.sub.n], y) [less than or equal to] [[mu].sub.n][phi]([x.sub.n], y) + 0(Sy, y) + 2<z - Sy, JSy - Jy)

This implies that

0 [less than or equal to] [phi](Sy, y) + 2<z - Sy, JSy - Jy>.

Using (2.4), we have that

0 [less than or equal to] [phi](Sy, y) + [phi](z, y) + [phi](Sy, Sy) - [phi](z Sy) - 0(Sy, y)

and hence [phi](z, Sy) [less than or equal to] [phi](z, y). This implies that z is an element of A(S). Similarly, we have that [phi](z, Ty) [less than or equal to] [phi](z, y) and hence z [member of] A(T). Therefore we have z [member of] A(S) [intersection] A(T). Additionally, if C is closed and convex and {[x.sub.n]} [subset] C, we have that z [member of] [bar.co]{[x.sub.n]: n [member of] N} [subset] C. In fact, if z [not member of] C, then there exists [y.sup.*] [member of] [E.sup.*] by the separation theorem [23] such that <z, [y.sup.*]> < [inf.sub.y [member of] C] <y,[y.sup.*]>. So, from {[x.sub.n]} [subset] C we have

[mathematical expression not reproducible].

This is a contradiction. Then we have z [member of] C. Since z [member of] A(S) [intersection] A(T) and z [member of] C, we have that

[phi](z, Sz) [less than or equal to] 0(z, z) = 0 and [phi](z, Tz) [less than or equal to] [phi](z, z) = 0

and hence [phi](z, Sz) = 0 and [phi](z, Tz) = 0. Since E is strictly convex, we have z [member of] F(S) [intersection] F(T). This completes the proof.

Using Lemma 3.1, we prove an attractive point theorem for commutative 2- generalized nonspreading mappings in a Banach space.

Theorem 3.2. Let C be a nonempty subset of a smooth, strictly convex and reflexive Banach space E and let S and T be commutative 2-generalized nonspreading mappings of C into itself. Suppose that there exists an element z [member of] C such that {[S.sup.k][T.sup.l]z: k, l [member of] N [union] {0}} is bounded. Then A(S) [intersection] A(T) is nonempty. Additionally, if C is closed and convex, then F(S) [intersection] F(T) is nonempty.

Proof. Since S is a 2-generalized nonspreading mapping of C into itself, there exist [mathematical expression not reproducible] such that for all x,y [member of] C,

(3.2) [mathematical expression not reproducible].

By assumption, we can take z [member of] C such that {[S.sup.k][T.sup.l]z : k,l [member of] N U {0}} is bounded. Replacing x by [S.sup.k][T.sup.l]z in (3.2), we have that for any y [member of] C and k, l [member of] N U {0},

[mathematical expression not reproducible].

This implies that

[mathematical expression not reproducible].

Summing up these inequalities with respect to k = 0,1, ..., n, we have

[mathematical expression not reproducible].

Furthermore, summing up these inequalities with respect to l = 0,1, ..., n, we have

[mathematical expression not reproducible].

Dividing by [(n + 1).sup.2], we have

[mathematical expression not reproducible],

where [mathematical expression not reproducible]. Since {[S.sup.k][T.sup.l]z} is bounded by assumption, {[S.sub.n]z} is bounded. Taking a Banach limit [mu] to both sides of this inequality, we have that

0 [less than or equal to] [phi](Sy, y) + 2[[mu].sub.n]<[S.sub.n]Z - Sy, JSy - Jy>

and hence

0 [less than or equal to] [phi](Sy, y) + [[mu].sub.n][phi]([S.sub.n]z, y) + [phi](Sy, Sy) - [[mu].sub.n][phi]([S.sub.n]z, Sy) - [phi](Sy, y)

Thus, we have

[[mu].sub.n][phi] ([S.sub.n] z, Sy) [less than or equal to] [[mu].sub.n] [phi] ([S.sub.n]z, y).

Similarly, replacing S and T by T and S, respectively, we have

[[mu].sub.n][phi]([S.sub.n]z, Ty) [less than or equal to] [[mu].sub.n][phi]([S.sub.n]z, y).

Using Lemma 3.1, we have that A(S) [intersection] A(T) is nonempty. Additionally, if C is closed and convex, then F(S) [intersection] F(T) is nonempty.

Since commutative 2-generalized hybrid mappings in a Hilbert space are commutative 2-generalized nonspreading mappings in a Banach space, as a direct sequence of Theorem 3.2, we have the following theorem proved by Hojo, Takahashi and Takahashi [6] in a Hilbert space.

Theorem 3.3 ([6]). Let H be a Hilbert space, let C be a nonempty subset of H and let S and T be commutative 2-generalized hybrid mappings of C into itself. Suppose that there exists an element z G C such that {[S.sup.k][T.sup.l]z: k,l [member of] NU{0}} is bounded. Then A(S) [intersection] A(T) is nonempty. Additionally, if C is closed and convex, then F(S) if F(T) is nonempty.

4. NONLINEAR ERGODIC THEOREMS

Let E be a smooth Banach space, let C be a nonempty subset of E and let J be the duality mapping from E into [E.sup.*]. Observe that if T : C [right arrow] E is a 2-generalized nonspreading mapping and F (T) [not equal to] 0, then

[phi](u,Ty) [less than or equal to] [phi](u, y)

for all u [member of] F(T) and y [member of] C. Indeed, putting x = u [member of] F(T) in (3.1), we obtain that

[mathematical expression not reproducible].

So, we have that

(4.1) [phi](u, Ty) [less than or equal to] [phi](u, y)

for all u [member of] F(T) and y [member of] C. Similarly, putting y = u [member of] F(T) in (3.1), we obtain that for x [member of] C,

[mathematical expression not reproducible]

and hence

[mathematical expression not reproducible].

If [mathematical expression not reproducible], then we have from (4.1) that

[mathematical expression not reproducible].

So, we have that

(4.2) [phi](Tx,u) [less than or equal to] [phi](x,u)

for all x [member of] C and u [member of] F(T). This implies that T is generalized nonexpansive in the sense of [8].

Now using the technique developed by [22] and [28], we can prove a mean convergence theorem without convexity for commutative 2-generalized nonspreading mappings in a Banach space. For proving this result, we need the following lemmas.

Lemma 4.1. Let E be a smooth, strictly convex and reflexive Banach space and let C be a nonempty subset of E. Let S and T be commutative 2-generalized nonspreading mappings of C into itself. If [[S.sup.k][T.sup.l]z : k,l [member of] N [union] {0}} for some z [member of] C is bounded and

[S.sub.n]x = 1/[(1 + n).sup.2] [n.summation over (k=0)] [n.summation over (l=0)] [S.sup.k][T.sup.l]x

for all x [member of] C and n [member of] N [union] {0}, then every weak cluster point of {[S.sub.n]x} is a point of A(S) [intersection] A(T). Additionally, if C is closed and convex, then every weak cluster point of {[S.sub.n]x} is a point of F(S) [intersection] F(T).

Proof. Since S : C [right arrow] C is 2-generalized nonspreading, we have that for all x,y [member of] C, (3.2) holds. Since there exists z [member of] C such that {[S.sup.k][T.sup.l]z : k,l [member of] N U {0}} is bounded, {[S.sup.k][T.sup.l]x : k,l [member of] NU{0}} for all x [member of] C is bounded. Then as in the proof of Theorem 3.2, we have that for any y [member of] C

[mathematical expression not reproducible].

Since {[S.sup.k][T.sup.l]x} is bounded, {[S.sub.n]x} is bounded. Thus we have a subsequence {[mathematical expression not reproducible]} of {[S.sub.n]x} such that {[mathematical expression not reproducible]} converges weakly to a point u [member of] E. Letting [n.sub.i] [right arrow] [infinity], we obtain

0 [less than or equal to] f(Sy, y) + 2<u - Sy, JSy - Jy>.

Using (2.4), we have that

0 [less than or equal to] [phi](Sy, y) + [phi](u, y) + [phi](Sy, Sy) - [phi](u, Sy) - [phi](Sy, y)

and hence

[phi](u, Sy) [less than or equal to] [phi](u, y)

This implies that u is an element of A(S). Similarly, we have that

[phi](u, Ty) [less than or equal to] [phi](u, y).

and hence u [member of] A(T). Therefore we have u [member of] A(S) [intersection] A(T). Additionally, if C is closed and convex, we have that {[S.sub.n]x} [subset] C and then

u [member of] [bar.co]{[S.sub.n]x : n [member of] N} [subset] C.

Since u [member of] A(S) [intersection] A(T) and u [member of] C, we have that

[phi](u, Su) [less than or equal to] [phi](u, u) = 0 and [phi](u, Tu) [less than or equal to] [phi](u, u) = 0

and hence

[phi](u, Su) = 0 and [phi](u,Tu) = 0.

Since E is strictly convex, we have [member of] G F(S) [intersection] F(T). This completes the proof.

Let E be a smooth Banach space. Let C be a nonempty subset of E and let T be a mapping of C into E. We denote by B(T) the set of skew-attractive points of T, i.e., B(T) = {z [member of] E: [phi](Tx, z) [less than or equal to] [phi](x, z), [for all]x [member of] C}. The following result was proved by Lin and Takahashi [20].

Lemma 4.2 ([20]). Let E be a smooth Banach space and let C be a nonempty subset of E. Let T be a mapping from C into E. Then B(T) is closed.

Let E be a smooth, strictly convex and reflexive Banach space and let C be a nonempty subset of E. Let T be a mapping of C into E. Define a mapping [T.sup.*] as follows:

[T.sup.*][x.sup.*] = JT[J.sup.-1][x.sup.*], V[x.sup.*] [member of] JC,

where J is the duality mapping on E and [J.sup.-1] is the duality mapping on [E.sup.*]. A mapping [T.sup.*] is called the duality mapping of T; see also [31] and [7]. It is easy to show that if T is a mapping of C into itselt, then [T.sup.*] is a mapping of JC into itself. In fact, for [x.sup.*] [member of] JC, we have [J.sup.-1] [x.sup.*] [member of] C and hence T[J.sup.-1] [x.sup.*] [member of] C. So, we have

[T.sup.*][x.sup.*] = JT[J.sup.-1] [x.sup.*] [member of] JC.

Then, [T.sup.*] is a mapping of JC into itself. Using Lemma 2.4, we have the following result.

Lemma 4.3 ([20]). Let E be a smooth, strictly convex and reflexive Banach space and let C be a nonempty subset of E. Let T be a mapping of C into E and let [T.sup.*] be the duality mapping of T. Then, the following hold:

(1) JB(T) = A([T.sup.*]);

(2) JA(T) = B([T.sup.*]).

In particular, JB(T) is closed and convex.

Let D = {(k,l) : k,l [member of] N [union] {0}}. Then D is a directed set by the binary relation:

(k,l) [less than or equal to] (i,j) if k [less than or equal to] i and l [less than or equal to] j.

Theorem 4.4. Let E be a uniformly convex Banach space with a Frechet differentiable norm and let C be a nonempty subset of E. Let S,T : C [right arrow] C be commutative 2-generalized nonspreading mappings such that {[S.sup.k][T.sup.l]z : k,l [member of] N [union] {0}} for some z [member of] C is bounded, A(S) = B(S) and A(T) = B(T). Let R be the sunny generalized nonexpansive retraction of E onto B(S) [intersection] B(T). Then, for any x [member of] C,

[S.sub.n]x = 1/[(n + 1).sup.2] [n.summation over (k=0)][N.summation over (l=0)] [S.sup.k][T.sup.l]x

converges weakly to an element q of A(S) [intersection] A(T), where q = [lim.sub.(k,l)[member of]D] R[S.sup.k][T.sup.l]x.

Proof. We have from Theorem 3.2 that A(S) [intersection] A(T) = B(S) [intersection] B(T) is nonempty. We know from Lemmas 2.11, 4.2 and 4.3 that B(S) [intersection] B(T) is closed, and

J(B(S) [intersection] B(T)) = JB(S) [intersection] JB(T)

is closed and convex. So, from Lemma 2.5 and Lemma 2.7 there exists the sunny generalized nonexpansive retraction R of E onto B(S) [intersection] B(T). From Lemma 2.8, this retraction R is characterized by

[mathematical expression not reproducible].

We also know from Lemma 2.6 that

0 [less than or equal to] <v - Rv, JRv - Ju>, [for all]u [member of] B(S) [intersection] B(T), v [member of] C.

Adding up [phi](Rv,u) to both sides of this inequality, we have

(4.3) [mathematical expression not reproducible].

Since [phi](Sz,u) [less than or equal to] [phi](z, u) and [phi](Tz, u) [less than or equal to] [phi](z, u) for any u [member of] B(S) [intersection] B(T) and z [member of] C, it follows that for any (k, l), (i, j) [member of] D with (k, l) [less than or equal to] (i, j),

[mathematical expression not reproducible].

Hence the net [phi]([S.sup.k][T.sup.l]x, R[S.sup.k][T.sup.l]x) is nonincreasing. Putting u = R[S.sup.k][T.sup.l]x and v = [S.sup.l][T.sup.j]x with (k,l) [less than or equal to] (i, j) in (4.3), we have from Lemma 2.3 that

[mathematical expression not reproducible],

where g is a strictly increasing, continuous and convex real-valued function with g(0) = 0. From the properties of g, {R[S.sup.k][T.sup.l]x} is a Cauchy net; see [19]. Therefore {R[S.sup.k][T.sup.l]x} converges strongly to a point q [member of] B(S) [intersection] B(T). Next, consider a fixed x [member of] C and an arbitrary subsequence [mathematical expression not reproducible] which converges weakly to a point v. From the proof of Lemma 4.1, we know that v [member of] A(S) [intersection] A(T) = B(S) [intersection] B(T). Rewriting the characterization of the retraction R, we have that for any u [member of] B(S) [intersection] B (T),

0 [less than or equal to] <[S.sup.k][T.sup.l]x - R[S.sup.k][T.sup.l]x, [JRS.sup.k][T.sup.l]x - Ju>

and hence

[mathematical expression not reproducible],

where K is an upper bound for [parallel][S.sup.k][T.sup.l]x - [RS.sup.k][T.sup.l]x[parallel]. Summing up these inequalities for k = 0,1, ..., n and l = 0,1, ..., n and dividing by [(n + 1).sup.2], we arrive to

[mathematical expression not reproducible],

where [mathematical expression not reproducible]. Letting [n.sub.i] [right arrow] [infinity] and remembering that J is continuous, we get

<v - q, Ju - Jq> [less than or equal to], 0.

This holds for any u [member of] B(S) [intersection] B(T). Therefore Rv = q. But because v [member of] B(S) [intersection] B(T), we have v = q. Thus the sequence {[S.sub.n]x} converges weakly to the point q [member of] A(S) [intersection] A(T).

Using Theorem 4.4, we obtain the following theorems.

Theorem 4.5. Let E be a uniformly convex Banach space with a Frechet differentiable norm. Let S,T : E [right arrow] E be commutative [mathematical expression not reproducible]-generalized nonspreading mappings such that [[alpha].sub.1] - [[beta].sub.1] = 0, [mathematical expression not reproducible], respectively.

Assume that {[S.sup.k][T.sup.l]z : k,l [member of] N [union] {0}} for some z [member of] C is bounded. Let R be the sunny generalized nonexpansive retraction of E onto F(S) [intersection] F(T). Then, for any x [member of] E,

[mathematical expression not reproducible]

converges weakly to an element q of F(S) [intersection] F(T), where q = [lim.sub.(k,l) [member of] D] [RS.sup.k][T.sup.l]x.

Proof. Since {[S.sup.k][T.sup.l]z: k, l [member of] N [union] {0}} for some z [member of] C is bounded, we have that A(S) [intersection] A(T) = F(S) [intersection] F(T) is nonempty. We also know that [[alpha].sub.2] > [[beta].sub.2] together with [[alpha].sub.1] - [[beta].sub.1] = 0, [[gamma].sub.1] [less than or equal to] [[delta].sub.1] and [[gamma].sub.2] [less than or equal to] [[delta].sub.2] implies that

[phi](Sx, u) [less than or equal to] [phi](x, u)

for all x [member of] E and u [member of] F(S). Similarly, [[alpha]'.sub.2] > [[beta]'.sub.2] together with [[alpha]'.sub.1] - [[beta]'.sub.1] = 0, [[gamma]'.sub.1] [less than or equal to] [[delta]'.sub.1] and [[gamma]'.sub.2] [less than or equal to] [[delta]'.sub.2] implies that

[phi](Tx, v) [less than or equal to] [less than or equal to] [phi](x, v)

for all x [member of] E and v [member of] F(T). Thus, we have that F(S) = B(S) and F(T) = B(T). Therefore, we have the desired result from Theorem 4.4.

Theorem 4.6 ([6]). Let H be a Hilbert space and let C be a nonempty subset of H. Let S and T be commutative 2-generalized hybrid mappings of C into itself such that {[S.sup.k][T.sup.l]z : k, l [member of] N [union] {0}} for some z [member of] C is bounded. Let P be the metric projection of H onto A(S) if A(T). Then, for any x [member of] C,

[S.sub.n]x = 1/[(n + 1).sup.2] [n.summation over (k=0)] [n.summation over (l=0)] [S.sup.k][T.sup.l]x

converges weakly to an element q of A(S) [intersection] A(T), where q = [lim.sub.(k,l)[member of]D] P[S.sup.k][T.sup.l]x. In particular, if C is closed and convex, {[S.sub.n]x} converges weakly to an element q of F(S) f F(T).

Proof. We have from Theorem 3.2 that A(S) [intersection] A(T) is nonempty. We also have that A(S) = B(S) and A(T) = B(T). Since A(S) [intersection] A(T) is a nonempty, closed and convex subset of H, there exists the metric projection of H onto A(S) [intersection] A(T). In a Hilbert space, the metric projection of H onto A(S) [intersection] A(T) is equivalent to the sunny generalized nonexpansive retraction of H onto A(S) [intersection] A(T). On the other hand, commutative 2-generalized hybrid mappings S, T : C [right arrow] C are commutative 2-generalized nonspreading mappings. So, we have the desired result from Theorem 4.6. Furthermore, if C is closed and convex, we have that q G F(S) [intersection] F(T) and then {[S.sub.n]x} converges weakly to q [member of] F(S) [intersection] F(T).

Remark We do not know whether a mean convergence theorem of Baillon's type for nonspreading mappings in a Banach space holds or not.

Acknowledgements. The first author was partially supported by Grant-in-Aid for Scientific Research No. 15K04906 from Japan Society for the Promotion of Science. The second and the third authors were partially supported by the grant MOST 105- 2115-M-037-001 and the grant MOST 105-2115-M-039-002-MY3, respectively.

REFERENCES

[1] Y. I. Alber, Metric and generalized projections in Banach spaces: Properties and applications, in Theory and Applications of Nonlinear Operators of Accretive and Monotone Type (A. G. Kartsatos Ed.), Marcel Dekker, New York, 1996, pp. 15-50.

[2] J.-B. Baillon, Un theoreme de type ergodique pour les contractions non lineaires dans un espace de Hilbert, C.R. Acad. Sci. Paris Ser. A-B 280:1511-1514, 1975.

[3] R. E. Bruck, Jr, A simple proof of the mean ergodic theorem for nonlinear contractions in Banach spaces, Israel J. Math. 32:107-116, 1979.

[4] K. Goebel and W. A. Kirk, Topics in Metric Fixed Point Theory, Cambridge University Press, Cambridge, 1990.

[5] N. Hirano, A proof of the mean ergodic theorem for nonexpansive mappings in Banach space, Proc. Amer. Math. Soc. 78:361-365, 1980.

[6] M. Hojo, S. Takahashi and W. Takahashi, Attractive point and ergodic theorems for two nonlinear mappings in Hilbert spaces, to appear.

[7] T. Honda, T. Ibaraki and W. Takahashi, Duality theorems and convergence theorems for nonlineaqr mappings in Banach spaces, Int. J. Math. Statis. 6:46-64, 2010.

[8] T. Ibaraki and W. Takahashi, A new projection and convergence theorems for the projections in Banach spaces, J. Approx. Theory 149:1-14, 2007.

[9] T. Ibaraki and W. Takahashi, Generalized nonexpansive mappings and a proximal-type algorithm in Banach spaces, in Nonlinear Analysis and Optimization I: Nonlinear Analysis, Contemporary Mathematics, Vol. 513, Amer. Math. Soc., Providence, RI, 2010, pp. 169-180.

[10] T. Igarashi, W. Takahashi and K. Tanaka, Weak convergence theorems for nonspreading mappings and equilibrium problems, in Nonlinear Analysis and Optimization (S. Akashi, W. Takahashi and T. Tanaka Eds.), Yokohama Publishers, Yokohama, 2008, pp. 75-85.

[11] S. Itoh and W. Takahashi, The common fixed point theory of single-valued mappings and multi-valued mappings, Pacific J. Math. 79:493-508, 1978.

[12] S. Kamimura and W. Takahashi, Strong convergence of a proximal-type algorithm in a Banach apace, SIAM J. Optim. 13:938-945, 2002.

[13] P. Kocourek, W. Takahashi and J.-C. Yao, Fixed point theorems and weak convergence theorems for generalized hybrid mappings in Hilbert spaces, Taiwanese J. Math. 14:2497- 2511, 2010.

[14] P. Kocourek, W. Takahashi and J.-C. Yao, Fixed point theorems and ergodic theorems for nonlinear mappings in Banach spaces, Adv. Math. Econ. 15:67-88, 2011.

[15] F. Kohsaka, Existence and approximation of common fixed points of two hybrid mappings in Hilbert spaces, J. Nonlinear Convex Anal. 16:2193-2205, 2015.

[16] F. Kohsaka and W. Takahashi, Generalized nonexpansive retractions and a proximal-type algorithm in Banach spaces, J. Nonlinear Convex Anal. 8:197-209, 2007.

[17] F. Kohsaka and W. Takahashi, Existence and approximation of fixed points of firmly nonexpansive-type mappings in Banach spaces, SIAM J. Optim. 19:824-835, 2008.

[18] F. Kohsaka and W. Takahashi, Fixed point theorems for a class of nonlinear mappings related to maximal monotone operators in Banach spaces, Arch. Math. 91:166-177, 2008.

[19] A. T. Lau and W. Takahashi, Weak convergence and nonlinear ergodic theorems for reversible semigroups of nonexpansive mappings, Pacific J. Math. 126:277-294, 1987.

[20] L.-J. Lin and W. Takahashi, Attractive point theorems for generalized nonspreading mappings in Banach spaces, J. Convex Anal. 20:265-284, 2013

[21] T. Maruyama, W. Takahashi and M. Yao, Fixed point and mean ergodic theorems for new nonlinear mappings in Hilbert spaces, J. Nonlinear Convex Anal. 12:185-197, 2011.

[22] W. Takahashi, A nonlinear ergodic theorem for an amenable semigroup of nonexpansive mappings in a Hilbert space, Proc. Amer. Math. Soc. 81:253-256, 1981.

[23] W. Takahashi, Nonlinear Functional Analysis, Yokohoma Publishers, Yokohoma, 2000.

[24] W. Takahashi, Convex Analysis and Approximation of Fixed Points (Japanese), Yokohama Publishers, Yokohama, 2000.

[25] W. Takahashi, Fixed point theorems for new nonlinear mappings in a Hilbert space, J. Nonlinea Convex Anal. 11:79-88, 2010.

[26] W. Takahashi, Weak convergence theorems for two generalized nonspreading mappings in Banach spaces, J. Nonlinea Convex Anal., to appear.

[27] W. Takahashi and Y. Takeuchi, Nonlinear ergodic theorem without convexity for generalized hybrid mappings in a Hilbert space, J. Nonlinear Convex Anal. 12:399-406, 2011.

[28] W. Takahashi and M. Toyoda, Weak convergence theorems for nonexpansive mappings and monotone mappings, J. Optim. Theory Appl. 118:417-428, 2003.

[29] W. Takahashi, N.-C Wong and J.-C. Yao, Fixed point theorems for three new nonlinear mappings in Banach spaces, J. Nonlinear Convex Anal. 13:363-381, 2012.

[30] W. Takahashi, N.-C Wong and J.-C. Yao, Fixed point theorems and convergence theorems for generalized nonspreading mappings in Banach spaces, J. Fixed Point Theory Appl. 11:159-183, 2012.

[31] W. Takahashi and J.-C. Yao, Nonlinear operators of monotone type and convergence theorems with equilibrium problems in Banach spaces, Taiwanese J. Math. 15:787-818, 2011.

[32] W. Takahashi and J.-C. Yao, Weak and strong convergence theorems for positively homogeneous nonexpansive mappings in Banach spaces, Taiwanese J. Math. 15:961-980, 2011.

[33] H. K. Xu, Inequalities in Banach spaces with applications, Nonlinear Anal. 16:1127-1138, 1981.

WATARU TAKAHASHI, CHING-FENG WEN, AND JEN-CHIH YAO

Dedicated to Professor Ravi Agarwal on the occasion of his 70th birthday

Center for Fundamental Science, Kaohsiung Medical University, Kaohsiung 80702, Taiwan; Keio Research and Education Center for Natural Sciences, Keio University, Kouhoku-ku, Yokohama 223-8521, Japan; and Department of Mathematical and Computing Sciences, Tokyo Institute of Technology, Ookayama, Meguro-ku, Tokyo 152-8552, Japan

wataru@is.titech.ac.jp; wataru@a00.itscom.net

Center for Fundamental Science, Kaohsiung Medical University, Kaohsiung 80702, Taiwan

cfwen@kmu.edu.tw

Center for General Education, China Medical University, Taichung 40402, Taiwan

yaoj c@mail.cmu.edu.tw