Professional Documents
Culture Documents
Variationaal Inequalities Approaches To Minimization Problems With Constraints of Generalized Mixed Eequilibria and Variational Inclusion
Variationaal Inequalities Approaches To Minimization Problems With Constraints of Generalized Mixed Eequilibria and Variational Inclusion
Article
Variational Inequalities Approaches to Minimization
Problems with Constraints of Generalized Mixed
Equilibria and Variational Inclusions
Lu-Chuan Ceng 1 and Mihai Postolache 2,3,4, * , Ching-Feng Wen 5 and Yonghong Yao 6,7
1 Department of Mathematics, Shanghai Normal University, Shanghai 200234, China; zenglc@hotmail.com
2 Center for General Education, China Medical University, Taichung 40402, Taiwan
3 Romanian Academy, Gh. Mihoc-C. Iacob Institute of Mathematical Statistics and Applied Mathematics,
050711 Bucharest, Romania
4 Department of Mathematics and Informatics, University “Politehnica” of Bucharest,
060042 Bucharest, Romania
5 Center for Fundamental Science, Kaohsiung Medical University, Kaohsiung 807, Taiwan;
cfwen@kmu.edu.tw
6 School of Mathematical Sciences, Tianjin Polytechnic University, Tianjin 300387, China;
yaoyonghong@aliyun.com
7 The Key Laboratory of Intelligent Information and Data Processing of NingXia Province,
North Minzu University, Yinchuan 750021, China
* Correspondence: mihai@mathem.pub.ro
Received: 18 February 2019; Accepted: 13 March 2019; Published: 16 March 2019
Abstract: Multistep composite implicit and explicit extragradient-like schemes are presented for
solving the minimization problem with the constraints of variational inclusions and generalized
mixed equilibrium problems. Strong convergence results of introduced schemes are given under
suitable control conditions.
1. Introduction
Let H be a real Hilbert space and ∅ 6= C ⊂ H be a closed convex set. Let A : C → H be a nonlinear
operator. Let us consider the variational inequality problem (VIP) which aims to find x ∗ ∈ C verifying
hA x ∗ , x − x ∗ i ≥ 0, ∀ x ∈ C. (1)
They showed that {uk } strongly converges to x † ∈ Fix( T ) which solves the following VIP
Ceng, Guu and Yao proved that { xt } and { xn } converge strongly to the same point x † ∈ Fix( T ),
which solves the following VIP
In [2], Peng and Yao considered the generalized mixed equilibrium problem (GMEP) which aims
to find x † ∈ C verifying
ϕ(y) − ϕ( x † ) + Θ( x † , y) + hA x † , y − x † i ≥ 0, ∀y ∈ C, (7)
( GEP) : Θ( x † , y) + hA x † , y − x † i ≥ 0, ∀y ∈ C,
( MEP) : Θ( x † , y) + ϕ(y) − ϕ( x † ) ≥ 0, ∀y ∈ C,
( EP) : Θ( x † , y) ≥ 0, ∀y ∈ C,
min f ( x ), (8)
x ∈C
where f : C → R is a convex and continuously Fréchet differentiable functional. Use Ξ to denote the
set of minimizers of CMP (8).
For solving (8), an efficient algorithm is the gradient-projection algorithm (GPA) which is defined
by: for given the initial guess x0 ,
or a general form,
zn+1 := PC (zk − µk ∇ f (zk )), ∀k ≥ 0. (10)
Mathematics 2019, 7, 270 3 of 20
0 ∈ Bx + Rx. (11)
2. Preliminaries
Let C be a nonempty closed convex subset of a real Hilbert space H. Recall that the metric
projection PC : H → C is defined by k x − PC x k = infy∈C k x − yk =: d( x, C ). A mapping A : H → H
is called strongly positive if h Ax, x i ≥ γ̄k x k2 , ∀ x ∈ H for γ̄ > 0. A mapping F : C → H is called
Lipschitz if k Fx − Fyk ≤ Lk x − yk, ∀ x, y ∈ C for some L ≥ 0. F is called nonexpansive if L = 1 and it
is called contractive provided L ∈ [0, 1).
• z = PC x ⇔ h x − z, z − x † i ≥ 0, ∀ x † ∈ C;
• z = PC x ⇔ k x − zk2 + k x † − zk2 ≤ k x − x † k2 , ∀ x † ∈ C;
• h PC x − PC x † , x − x † i ≥ k PC x − PC x † k2 , ∀ x † ∈ H.
• monotone if h Fx − Fx † , x − x † i ≥ 0, ∀ x, x † ∈ C;
• η-strongly monotone if h Fx − Fx † , x − x † i ≥ η k x − x † k2 , ∀ x, y ∈ C for some η > 0;
• α-inverse-strongly monotone (ism) if h Fx − Fx † , x − x † i ≥ αk Fx − Fx † k2 , ∀ x, x † ∈ C for some α > 0.
In this case, we have for all u, u† ∈ C,
Let ϕ : C → R be a lower semicontinuous and convex function satisfying the following conditions:
(B1) for each x ∈ H and r > 0, there exists a bounded subset Dx ⊂ C and y x ∈ C such that for any
z ∈ C \ Dx ,
1
Θ(z, y x ) + ϕ(y x ) − ϕ(z) + hy x − z, z − x i < 0;
r
(B2) C is a bounded set.
(Θ,ϕ)
Let r > 0. Define a mapping Tr : H → C by the following form: for any x ∈ H,
(Θ,ϕ) 1
Tr ( x ) := { x † ∈ C : Θ( x † , y) + ϕ(y) − ϕ( x † ) + h x † − x, y − x † i ≥ 0, ∀y ∈ C }.
r
If ϕ ≡ 0, then TrΘ ( x ) := { x † ∈ C : Θ( x † , y) + 1r h x † − x, y − x † i ≥ 0, ∀y ∈ C }.
k x + x † k2 ≤ k x k2 + 2h x † , x + x † i, ∀ x, x † ∈ E.
• k x − x † k2 = k x k2 − 2h x − x † , x † i − k x † k2 , ∀ x, x † ∈ H;
• kλx + γx † k2 = λk x k2 + γk x † k2 − λγk x − x † k2 , ∀ x, x † ∈ H and λ, γ ∈ [0, 1] with λ + γ = 1;
• {un } ⊂ H is a sequence satisfying un * x. Then,
Lemma 3 ([45]). Let H be a real Hilbert space and ∅ 6= C ⊂ H be a closed convex set. Let T : C → C be a
k-strictly pseudocontractive mapping. Then,
• k Tx − Tyk ≤ 11+ k
−k k x − y k,∀ x, y ∈ C.
• I − T is demiclosed at 0.
• Fix( T ) of T is closed and convex.
Lemma 4 ([39]). Let H be a real Hilbert space and ∅ 6= C ⊂ H be a closed convex set. Let T : C → C be a
k-strictly pseudocontractive mapping. Then,
kγ( x − x † ) + µ( Tx − Tx † )k ≤ (γ + µ)k x − x † k, ∀ x, x † ∈ C,
Mathematics 2019, 7, 270 5 of 20
2η
Lemma 5 ([42]). Let 0 < µ < κ2 . Then, for ∀ x, y ∈ C, we have k T λ x − T λ yk ≤ (1 − λτ )k x − yk,
p
where τ = 1 − 1 − µ(2η − µκ 2 ).
Lemma 6 ([46]). Let { an } ⊂ [0, +∞), {ωn } ⊂ [0, 1], {δn } and {rn } ⊂ [0, +∞) be four sequences. If
(i) ∑∞ ∞
n=0 ωn = ∞ and ∑n=1 rn < ∞;
(ii) either lim supn→∞ δn ≤ 0 or ∑∞ n=0 ωn | δn | < ∞;
(iii) an+1 ≤ (1 − ωn ) an + ωn δn + rn , ∀n ≥ 0.
Then, limn→∞ an = 0.
Lemma 7 ([44]). Let bounded linear operator A : H → H be γ̄-strongly positive. Then, k I − ρAk ≤ 1 − ργ̄
provided 0 < ρ ≤ k Ak−1 .
• an ≤ cn , n ≥ 1⇒LIMn an ≤ LIMn cn ;
• LIMn an+ N = LIMn an , where N is a fixed positive integer;
• lim infn→∞ an ≤ LIMn an ≤ lim supn→∞ an , ∀{ an } ∈ l ∞ .
Lemma 8 ([40]). Assume that the sequence {sn } ∈ l ∞ satisfy LIMn sn ≤ M, where M is a constant.
If lim supn→∞ (sn+1 − sn ) ≤ 0, then lim supn→∞ sn ≤ M.
1 1
k JR,λ x − JR,µ x † k ≤ k x − x † k + |λ − µ|( k JR,λ x − x † k + k x − JR,µ x † k), ∀ x, x † ∈ H. (13)
λ µ
3. Main Results
Let H be a real Hilbert space and ∅ 6= C ⊂ H be a closed convex set. In what follows, we assume:
f : C → R is a convex functional with gradient ∇ f being L-Lipschitz, the operator F : C → H
is η-strongly monotone and κ-Lipschitzian, Ri : C → 2 H is a maximal monotone mapping and
Bi : C → H (i = 1, ..., N ) is ηi -ism, Θ j : C × C → R satisfies (A1)–(A4), ϕ j : C → R ∪ {+∞} is a
proper convex and lower semicontinuous satisfying (B1) or (B2), and A j : C → H is µ j -ism for each
j = 1, ..., M;
V : C → H is a nonexpansive mapping and A : H → H is a γ̄-strongly positive bounded linear
2η p
operator such 1 < γ̄ < 2, 0 < µ < κ2 and 0 ≤ γ < τ with τ = 1 − 1 − µ(2η − µκ 2 );
Mathematics 2019, 7, 270 6 of 20
Λit = JRi ,λi,t ( I − λi,t Bi ) JRi−1 ,λi−1,t ( I − λi−1,t Bi−1 ) · · · JR1 ,λ1,t ( I − λ1,t B1 ), ∀t ∈ (0, 1),
Λin = JRi ,λi,n ( I − λi,n Bi ) JRi−1 ,λi−1,n ( I − λi−1,n Bi−1 ) · · · JR1 ,λ1,n ( I − λ1,n B1 ), ∀n ≥ 0,
j (Θ j ,ϕ j ) (Θ j −1 ,ϕ j−1 ) (Θ ,ϕ1 )
∆t = Tr j,t ( I − r j,t A j ) Tr j−1,t ( I − r j−1,t A j−1 ) · · · Tr1,t 1 ( I − r1,t A1 ), ∀t ∈ (0, 1),
and
j (Θ j ,ϕ j ) (Θ j −1 ,ϕ j−1 ) (Θ ,ϕ1 )
∆n = Tr j,n ( I − r j,n A j ) Tr j−1,n ( I − r j−1,n A j−1 ) · · · Tr1,n1 ( I − r1,n A1 ), ∀n ≥ 0,
On the other hand, since {ri,t } ⊂ [ci , di ] ⊂ (0, 2µi ), according to (12) and Proposition 2, we get
(Θ ,ϕ M ) (Θ ,ϕ M )
k∆tM x − ∆tM yk = k Tr M,tM ( I − r M,t A M )∆tM−1 x − Tr M,tM ( I − r M,t A M )∆tM−1 yk
≤ k( I − r M,t A M )∆tM−1 x − ( I − r M,t A M )∆tM−1 yk
≤ k x − y k.
(Θ M ,ϕ M ) (Θ (Θ ,ϕ )
,ϕ )
ut = Tr M,t
( I − r M,t A M ) Tr M−M1,t−1 M−1 ( I − r M−1,t A M−1 ) · · · Tr1,t 1 1 ( I − r1,t A1 ) xt ,
vt = JR N ,λ N,t ( I − λ N,t BN ) JR N −1 ,λ N −1,t ( I − λ N −1,t BN −1 ) · · · JR1 ,λ1,t ( I − λ1,t B1 )ut , (14)
x = P [( I − s A) T v + s [Vx − t(µFVx − γT v )].
t C t t t t t t t t
h( A − V ) x̃, p − x̃ i ≥ 0, ∀ p ∈ Ω. (15)
Mathematics 2019, 7, 270 7 of 20
(Θ ,ϕ ) (Θ −1 ,ϕ M−1 ) (Θ ,ϕ )
un = Tr M,nM M ( I − r M,n A M ) Tr M−M1,n ( I − r M−1,n A M−1 ) · · · Tr1,n1 1 ( I − r1,n A1 ) xn ,
R N ,λ N,n ( I − λ N,n B N ) JR N −1 ,λ N −1,n ( I − λ N −1,n B N −1 ) · · · JR1 ,λ1,n ( I − λ1,n B1 ) un ,
v = J
n
(16)
yn = αn γTn vn + ( I − αn µF )Vxn ,
n+1 = PC [( I − sn A ) Tn vn + sn yn ], ∀ n ≥ 0.
x
It is easy to check that 0 < 1 − st (γ̄ − 1 + t(τ − γ)) < 1. Therefore, Qt : C → C is contraction and
Qt has a unique fixed point, denoted by xt .
j
Proof. (i) Pick up p ∈ Ω. Noting that Fix( Tt ) = Ξ, Λit p = p and ∆t p = p for all i ∈ {1, ..., N } and
j
j ∈ {1, ..., M }, by the nonexpansivity of Tt , Λit and ∆t and Lemmas 5 and 7 we get
k xt − pk ≤ k( I − st A) Tt ΛtN ∆tM xt − ( I − st A) Tt ΛtN ∆tM pk + st k( I − tµF )Vxt + tγTt ΛtN ∆tM xt − Apk
≤ st k( I − tµF )Vxt − ( I − tµF )V p + t(γTt ΛtN ∆tM xt − µFV p) + V p − Apk
+ (1 − st γ̄)k Tt ΛtN ∆tM xt − Tt ΛtN ∆tM pk
≤ [1 − st (γ̄ − 1 + t(τ − γ))]k xt − pk + st (tk(γI − µFV ) pk + k(V − A) pk).
Thus,
k(V − A) pk + k(γI − µFV ) pk
k xt − pk ≤ .
γ̄ − 1
Hence { xt } is bounded and so are {Vxt }, {ut }, {vt }, { Tt vt } and { FVxt }.
(ii) From (14), we obtain
and
j
k u t − p k2 ≤ k ∆ t x t − p k2
(Θ ,ϕ j ) j −1 (Θ j ,ϕ j )
= k Tr j,t j ( I − r j,t A j )∆t xt − Tr j,t ( I − r j,t A j ) pk2 (19)
j −1
≤ k xt − pk2 + r j,t (r j,t − 2µ j )k A j ∆t x t − A j p k2 .
Simple calculations show that
which together with {λi,t } ⊂ [ ai , bi ] ⊂ (0, 2ηi ), i = 1, ..., N and {r j,t } ⊂ [c j , d j ] ⊂ (0, 2µ j ), j = 1, ..., M,
implies that
1 − st (γ̄ − tγ) j −1
[c j (2µ j − d j )k A j ∆t xt − A j pk2 + ai (2ηi − bi )k Bi Λit−1 ut − Bi pk2 ]
2
1 − st (γ̄ − tγ) j −1
≤ [r j,t (2µ j − r j,t )k A j ∆t xt − A j pk2 + λi,t (2ηi − λi,t )k Bi Λit−1 ut − Bi pk2 ]
2
≤ st (tk(γI − µFV ) pk + k(V − A) pk)k xt − pk.
j j −1 j
k∆t xt − pk2 ≤ h( I − r j,t A j )∆t xt − ( I − r j,t A j ) p, ∆t xt − pi
1 j −1 j j −1 j j −1
≤ (k∆t xt − pk2 + k∆t xt − pk2 − k∆t xt − ∆t xt − r j,t ( A j ∆t xt − A j p)k2 ),
2
which implies that
j j −1 j −1 j j −1
k ∆ t x t − p k2 ≤ k ∆ t x t − p k2 − k ∆ t xt − ∆t xt − r j,t ( A j ∆t xt − A j p)k2
j −1 j j −1 j j −1
(23)
≤ k x t − p k2 − k ∆ t xt − ∆t xt k2 + 2r j,t k∆t xt − ∆t xt kk A j ∆t x t − A j p k.
1
k xt − pk2 ≤ (1 − st (γ̄ − tγ)) (kΛit ut − pk2 + k xt − pk2 ) + st k(V − A) pkk xt − pk
2
+ st [(1 − tτ )k xt − pk2 + tk(γI − µFV ) pkk xt − pk]
1 j −1 j
≤ (1 − st (γ̄ − tγ)) [k xt − pk2 − k∆t xt − ∆t xt k2
2
j −1 j j −1
+ 2r j,t k∆t xt − ∆t xt kk A j ∆t xt − A j pk − kΛit−1 ut − Λit ut k2
+ 2λi,t kΛit−1 ut − Λit ut kk Bi Λit−1 ut − Bi pk + k xt − pk2 ]
+ st [(1 − tτ )k xt − pk2 + tk(γI − µFV ) pkk xt − pk] + st k(V − A) pkk xt − pk
1 − st (γ̄ − tγ) j −1 j
≤ k x t − p k2 − (k∆t xt − ∆t xt k2 + kΛit−1 ut − Λit ut k2 )
2
j −1 j j −1
+ r j,t k∆t xt − ∆t xt kk A j ∆t xt − A j pk + λi,t kΛit−1 ut − Λit ut kk Bi Λit−1 ut − Bi pk
+ st (tk(γI − µFV ) pk + k(V − A) pk)k xt − pk,
which together with {λi,t } ⊂ [ ai , bi ] ⊂ (0, 2ηi ) and {r j,t } ⊂ [c j , d j ] ⊂ (0, 2µ j ), leads to
1 − st (γ̄ − tγ) j −1 j
(k∆t xt − ∆t xt k2 + kΛit−1 ut − Λit ut k2 )
2
j −1 j j −1
≤ d j k∆t xt − ∆t xt kk A j ∆t xt − A j pk + bi kΛit−1 ut − Λit ut kk Bi Λit−1 ut − Bi pk
+ st (tk(γI − µFV ) pk + k(V − A) pk)k xt − pk.
j −1
Since limt→0 st = 0, limt→0 k A j ∆t xt − A j pk = 0 and limt→0 k Bi Λit−1 ut − Bi pk = 0 (due to
j
(22)), we deduce from the boundedness of { xt }, {Λit ut } and {∆t xt } that
j −1 j
lim kΛit−1 ut − Λit ut k = 0 and lim k∆t xt − ∆t xt k = 0. (25)
t →0 t →0
Hence we get
and
kut − vt k ≤ kΛ0t ut − Λ1t ut k + kΛ1t ut − Λ2t ut k + · · · + kΛtN −1 ut − ΛtN ut k
(27)
→ 0 as t → 0.
So, taking into account that k xt − vt k ≤ k xt − ut k + kut − vt k, we have
lim k xt − vt k = 0. (28)
t →0
In the meantime, from the nonexpansivity of Tt and ΛtN , it is easy to see that
k xt − Tt xt k ≤ k xt − Tt vt k + k Tt vt − Tt xt k
≤ k xt − Tt vt k + kvt − xt k,
and
k xt − ΛtN xt k ≤ k xt − ΛtN ut k + kΛtN ut − ΛtN xt k
≤ k x t − v t k + k u t − x t k.
Consequently, from (17), (26) and (28) we immediately deduce that
k PC ( I − λt ∇ f )vt0 k ≤ k PC ( I − λt ∇ f )vt0 − PC ( I − λt ∇ f ) pk + k pk
≤ k v t0 k + 2 k p k .
where supt { Lk PC ( I − λt ∇ f )vt0 k + 4k∇ f (vt0 )k + Lkvt0 k} ≤ M e for some Me > 0. So, by (30),
we have that
k Tt vt − Tt0 vt0 k ≤ k Tt vt − Tt vt0 k + k Tt vt0 − Tt0 vt0 k
≤ k v t − v t0 k + Me |λt − λt |
0 (31)
4M e
≤ k v t − v t0 k + | s t − s t0 |.
L
Utilizing (12) and (13), we obtain that
Mathematics 2019, 7, 270 11 of 20
where
1
sup{ k JRi ,λi,t ( I − λi,t0 Bi )Λit−1 ut − ( I − λi,t0 Bi )Λit−
0
1
u t0 k
t λ i,t
1
+ k( I − λi,t0 Bi )Λit−1 ut − JRi ,λi,t ( I − λi,t0 Bi )Λit− 0
1
ut0 k} ≤ M,
b
λi,t0 0
M
j −1 1 (Θ ,ϕ ) j −1 j −1
∑ [k Bj ∆t xt k +
r j,t
k Tr j,t j j ( I − r j,t Bj )∆t xt − ( I − r j,t Bj )∆t xt k] ≤ M
e 1.
j =1
M N
4M
∑ |r j,t − r j,t0 | + Me 0 ∑ |λi,t − λi,t0 | +
e
k Tt vt − Tt0 vt0 k ≤ k xt − xt0 k + M
e1 | s t − s t0 |,
j =1 i =1
L
and hence
Mathematics 2019, 7, 270 12 of 20
4M e
+ |st − st0 |) + |st − st0 |[k Akk Tt vt k + kVxt k + t(γk Tt vt k + µk FVxt k)]
L
+ st0 (γk Tt vt k + µk FVxt k)|t − t0 |.
Since st , λi,t , r j,t are locally Lipschitzian, we conclude that xt is locally Lipschitzian.
Hence,
1
k x t − p k2 ≤ (h(V − A) p, xt − pi + th(γI − µFV ) p, xt − pi). (34)
γ̄ − 1 + t(τ − γ)
2− λ t n L
where stn = 4 ∈ (0, 21 ) for λtn ∈ (0, L2 ). Hence we have
2 2
k PC ( I − ∇ f ) xtn − xtn k ≤ k( I − ∇ f ) xtn − ( I − λtn ∇ f ) xtn k + k PC ( I − λtn ∇ f ) xtn − xtn k
L L
2
≤ ( − λtn )k∇ f ( xtn )k + k Ttn xtn − xtn k.
L
By the boundedness of { xtn }, stn → 0 (⇔ λtn → L2 ) and k Ttn xtn − xtn k → 0, we deduce
2 2
k x ∗ − PC ( I − ∇ f ) x ∗ k = lim k xtn − PC ( I − ∇ f ) xtn k = 0.
L n → ∞ L
Therefore, x ∗ ∈ VI(C, ∇ f ) = Ξ.
j
Furthermore, from (25), (26) and (28), we have that ∆tn xtn → x ∗ , Λm ∗
tn utn → x , u tn → x
∗
−1 m −1
Λm m
tn utn − λm,tn Bm Λtn utn ∈ ( I + λm,tn Rm ) Λtn utn ,
that is,
1
( Λ m −1 u t n − Λ m m −1 m
tn utn − λm,tn Bm Λtn utn ) ∈ Rm Λtn utn .
λm,tn tn
According to the monotonicity of Rm , we have
1 −1 m −1
hv − Λm
tn utn , g − Bm v − (Λm m
tn utn − Λtn utn − λm,tn Bm Λtn utn )i ≥ 0
λm,tn
and hence
1
hv − Λm m
tn utn , g i ≥ h v − Λtn utn , Bm v + ( Λ m −1 u t n − Λ m m −1
tn utn − λm,tn Bm Λtn utn )i
λm,tn tn
m −1 1
≥ hv − Λm m m
tn utn , Bm Λtn utn − Bm Λtn utn i + h v − Λtn utn , ( Λ m −1 u t n − Λ m
tn utn )i.
λm,tn tn
m −1 m −1
Since kΛmtn utn − Λtn utn k → 0 (due to (25)) and k Bm Λtn utn − Bm Λtn utn k → 0, we deduce from
m
Λm ∗
tn utn → x and { λm,tn } ⊂ [ am , bm ] ⊂ (0, 2ηm ) that
∗
lim hv − Λm
tn utn , g i = h v − x , g i ≥ 0.
n→∞
j j −1
j j j −1 j j ∆ tn xtn − ∆ tn x tn
Θ j ( ∆ tn xtn , y ) + ϕ j ( y ) − ϕ j ( ∆tn x tn ) + h A j ∆ tn xtn , y − ∆tn x tn i + h y − ∆ tn xtn , i ≥ 0.
r j,tn
By (A2), we have
j j −1
j j −1 j j ∆ tn x tn − ∆ tn xtn j
ϕ j ( y ) − ϕ j ( ∆tn x tn ) + h A j ∆ tn x tn , y − ∆tn xtn i + h y − ∆ tn xtn , i ≥ Θ j (y, ∆tn xtn ).
r j,tn
Mathematics 2019, 7, 270 14 of 20
j j j j j
h z t − ∆ tn x tn , A j zt i ≥ ϕ j ( ∆tn xtn ) − ϕ j ( z t ) + h z t − ∆ tn xtn , A j zt − A j ∆ tn xtn i + Θ j ( zt , ∆ tn xtn )
j j −1 (35)
j j j −1 j ∆ tn x tn − ∆ tn xtn
+ h zt − ∆ tn xtn , A j ∆ tn xtn − A j ∆ tn xtn i − h zt − ∆ tn xtn , i.
r j,tn
j j −1 j
By (25), we have k A j ∆tn xtn − A j ∆tn xtn k → 0 as n → ∞. Please note that hzt − ∆tn xtn , A j zt −
j
A j ∆tn xtn i ≥ 0. Then,
h z t − x ∗ , A j z t i ≥ ϕ j ( x ∗ ) − ϕ j ( z t ) + Θ j ( z t , x ∗ ). (36)
and hence
0 ≤ Θ j (zt , y) + ϕ j (y) − ϕ j (zt ) + (1 − t)hy − x ∗ , A j zt i.
Thus,
0 ≤ Θ j ( x ∗ , y ) + ϕ j ( y ) − ϕ j ( x ∗ ) + h y − x ∗ , A j x ∗ i.
Next, we prove that xt → x̃ as t → 0. First, let us assert that x ∗ is a solution of the VIP (15). As a
matter of fact, since xt = xt − wt + ( I − st A) Tt ΛtN ∆tM xt + st (( I − tµF )Vxt + tγTt ΛtN ∆tM xt ), we have
xt − Tt ΛtN ∆tM xt = xt − wt + st (V − A) Tt ΛtN ∆tM xt + st (Vxt − VTt ΛtN ∆tM xt + t(γTt ΛtN ∆tM xt − µFVxt )).
Since Tt , ΛtN and ∆tM are nonexpansive mappings, I − Tt ΛtN ∆tM is monotone. By the monotonicity of
I − Tt ΛtN ∆tM , we have
Hence,
Now, replacing t in (37) with tn and noticing the boundedness of {γk Ttn vtn k + µk FVxtn k} and
the fact that (V − A) Ttn vtn − (V − A) xtn → 0, we deduce
h( A − V ) x ∗ , x ∗ − pi ≤ 0.
Theorem 2. Assume that the sequences {αn } ⊂ [0, 1] and {sn } ⊂ (0, 12 ) satisfy: (C1) limn→∞ αn = 0
and limn→∞ sn = 0. Let the sequence { xn } be defined by (16). Then LIMn h( A − V ) x̃, x̃ − xn i ≤ 0,
where x̃ = limt→0+ xt and xt is defined by
Proof. We assume, without loss of generality, that 0 < sn ≤ k Ak−1 for all n ≥ 0. Let limt→0 xt =
x̃ and x̃ is the unique solution of (15). By Proposition 4, we deduce that the nets { xt },
{Vxt }, {∆ M xt }, {Λ N ∆ M xt } and { FVxt } are bounded.
Let p ∈ Ω. Then Tn p = p, ΛnN p = p and ∆nM p = p. Applying Lemmas 5 and 7, we obtain that
By induction
Thus, { xn }, {un }, {vn }, { Tn vn }, { FVxn }, {Vxn } and {yn } are all bounded. By (C1), we obtain
2
k TΛ N ∆ M xn − Tn ΛnN ∆nM xn k ≤ kΛ N ∆ M xn − ΛnN ∆nM xn k + M
b| − λ n |, (39)
L
where
1
sup{ kJ ( I − λi,n Bi )Λi−1 ∆ M xn − ( I − λi,n Bi )Λin−1 un k
n ≥0 λi Ri ,λi
1
+ k( I − λi,n Bi )Λi−1 ∆ M xn − JRi ,λi,n ( I − λi,n Bi )Λin−1 un k} ≤ N,
b
λi,n
N i −1 ∆ M x k + N
n≥0 { ∑i =1 k Bi Λ
b > 0 and sup
for some N n b} ≤ M b 0 > 0. In terms of (33),
b 0 for some M
we have
M
k∆ M xn − ∆nM xn k ≤ M
b1 ∑ |r j − r j,n |, (41)
j =1
Mathematics 2019, 7, 270 16 of 20
(Θ j ,ϕ j )
where supn≥0 {∑ jM=1 [k A j ∆ j−1 xn k + r1 k Tr j ( I − r j A j )∆ j−1 xn − ( I − r j A j )∆ j−1 xn k]} ≤ M
b 1 for some
j
b 1 > 0. In terms of (39)–(41) we calculate
M
M N
2
k TΛ N ∆ M xn − Tn ΛnN ∆nM xn k ≤ M
b1 ∑ |r j − r j,n | + Mb 0 ∑ |λi − λi,n | + Mb | L − λn |.
j =1 i =1
It is clear that
+ k Tn vn − xn+1 k
= k x t − x n k + en ,
Since
st h A( xt − xn ), xt − xn i ≤ st k Akk xt − xn k2 ≤ st k AkK2 → 0 as t → 0, (48)
Theorem 3. Let the sequences {αn } ⊂ [0, 1] and {sn } ⊂ (0, 12 ) satisfy: (C1) limn→∞ αn = 0, limn→∞ sn =
0 and (C2) ∑∞ n=0 sn = ∞. Let the sequence { xn } be defined by (16). If xn+1 − xn * 0, then xn converges
strongly to x̃ ∈ Ω, which solves (15).
2s (γ̄−1)
Proof. We assume, without loss of generality, that αn τ < 1 and n1−sn < 1 for all n ≥ 0. Let xt be
defined by (38). Then, limt→0 xt := x̃ ∈ Ω (due to Theorem 1). We divide the rest of the proof into
several steps.
Step 1. It is easy to show that
Hence { xn }, {un }, {vn }, { Tn vn }, { FVxn }, {Vxn } and {yn } are all bounded.
Step 2. We show that lim supn→∞ h(V − A) x̃, xn − x̃ i ≤ 0. To this end, let
an := h( A − V ) x̃, x̃ − xn i, ∀n ≥ 0.
Mathematics 2019, 7, 270 18 of 20
and therefore
2
δn = [αn k(γI − µFV ) x̃ kk xn+1 − x̃ k + h( A − V ) x̃, x̃ − xn+1 i].
γ̄ − 1 + αn (τ − γ)
Author Contributions: All the authors have contributed equally to this paper. All the authors have read and
approved the final manuscript.
Funding: This research was partially supported by the Innovation Program of Shanghai Municipal Education
Commission (15ZZ068), Ph.D. Program Foundation of Ministry of Education of China (20123127110002) and
Program for Outstanding Academic Leaders in Shanghai City (15XD1503100).
Conflicts of Interest: The authors declare no conflict of interest.
Mathematics 2019, 7, 270 19 of 20
25. Yao, Y.; Leng, L.; Postolache, M.; Zheng, X. Mann-type iteration method for solving the split common fixed
point problem. J. Nonlinear Convex Anal. 2017, 18, 875–882.
26. Thakur, B.-S.; Postolache, M. Existence and approximation of solutions for generalized extended nonlinear
variational inequalities. J. Inequal. Appl. 2013, 2013, 590.
27. Dadashi, V.; Postolache, M. Forward-backward splitting algorithm for fixed point problems and zeros of the
sum of monotone operators. Arab. J. Math. 2019, doi:10.1007/s40065-018-0236-2.
28. Dong, Q.-L.; Lu, Y.-Y.; Yang, J. The extragradient algorithm with inertial effects for solving the variational
inequality. Optimization 2016, 65, 2217–2226.
29. Zhao, X.-P.; Ng, K.-F.; Li, C.; Yao, J.-C. Linear regularity and linear convergence of projection-based methods
for solving convex feasibility problems, Appl. Math. Optim. 2018, 78, 613–641.
30. Yao, Y.-H.; Postolache, M.; Liou, Y.-C. Strong convergence of a self-adaptive method for the split feasibility
problem. Fixed Point Theory Appl. 2013, 2013, 201.
31. Yao, Y.; Yao, J.-C.; Liou, Y.-C.; Postolache, M. Iterative algorithms for split common fixed points of
demicontractive operators without priori knowledge of operator norms. Carpathian J. Math. 2018, 34, 459–466.
32. Yao, Y.; Liou, Y.-C.; Postolache, M. Self-adaptive algorithms for the split problem of the demicontractive
operators. Optimization 2018, 67, 1309–1319.
33. Rockafellar, R.-T. Monotone operators and the proximal point algorithms. SIAM J. Control Optim. 1976,
14, 877–898.
34. Huang, N.J. A new completely general class of variational inclusions with noncompact valued mappings.
Comput. Math. Appl. 1998, 35, 9–14.
35. Zeng, L.-C.; Guu, S.-M.; Yao, J.C. Characterization of H-monotone operators with applications to variational
inclusions. Comput. Math. Appl. 2005, 50, 329–337.
36. Ceng, L.-C.; Ansari, Q.-H.; Wong, M.-M.; Yao, J.-C. Mann type hybrid extragradient method for variational
inequalities, variational inclusions and fixed point problems. Fixed Point Theory 2012, 13, 403–422.
37. Ceng, L.-C.; Guu, S.-M.; Yao, J.-C. Iterative algorithm for finding approximate solutions of mixed
quasi-variational-like inclusions. Comput. Math. Appl. 2008, 56, 942–952.
38. Jung, J.-S. A general composite iterative method for strictly pseudocontractive mappings in Hilbert spaces.
Fixed Point Theory Appl. 2014, 2014, 173.
39. Yao, Y.; Liou, Y.-C.; Kang, S.-M. Approach to common elements of variational inequality problems and fixed
point problems via a relaxed extragradient method. Comput. Math. Appl. 2010, 59, 3472–3480.
40. Shioji, N.; Takahashi, W. Strong convergence of approximated sequences for nonexpansive mappings in
Banach spaces. Proc. Am. Math. Soc. 1997, 125, 3641–3645.
41. Byrne, C. A unified treatment of some iterative algorithms in signal processing and image reconstruction.
Inverse Probl. 2004, 20, 103–120.
42. Xu, H.-K.; Kim, T.-H. Convergence of hybrid steepest-descent methods for variational inequalities. J. Optim.
Theory Appl. 2003, 119, 185–201.
43. Barbu, V. Nonlinear Semigroups and Differential Equations in Banach Spaces; Springer:
Heidelberg, Germany, 1976.
44. Marino, G.; Xu, H.-K. A general iterative method for nonexpansive mappings in Hilbert spaces. J. Math.
Anal. Appl. 2006, 318, 43–56.
45. Marino, G.; Xu, H.-K. Weak and strong convergence theorems for strict pseudo-contractions in Hilbert
spaces. J. Math. Math. Appl. 2007, 329, 336–346.
46. Xu, H.-K. Iterative algorithms for nonlinear operators. J. Lond. Math. Soc. 2002, 66, 240–256.
c 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).