The concept of Levitin-Polyak well-posedness of an equilibrium-like problem in Banach spaces is introduced. Under suitable conditions, some characterizations of its Levitin-Polyak well-posedness are established. Some conditions under which an equilibrium-like problem in Banach spaces is Levitin-Polyak well-posed are also derived.
1. Introduction
In 1966, Tykhonov [1] first established the well-posedness of a minimization problem, which has been known as Tykhonov well-posedness. Since it is important in optimization problems, various concepts of well-posedness have been introduced and studied in past decades. For more about the well-posedness, we refer to [2–4] and the references therein.
The Tykhonov well-posedness of a constrained minimization problem requires that every minimizing sequence should lie in the constraint set. In many situations, the minimizing sequence produced by a numerical optimization method usually fails to be feasible but gets closer and closer to the constraint set. Levitin and Polyak [5] generalized the concept of Tykhnov well-posedness by requiring the existence and uniqueness of minimizer and the convergence of every generalized minimizing sequence toward the unique minimizer, which has been known as Levitin and Polyak well-posedness. There are a lot of results concerned with Tykhonov well-posedness, LP well-posedness, and their generalizations for minimization problems. For details, we refer to [1–3, 5–7].
Recently, the concept of well-posedness has been extended to many other fields, including Nash equilibrium [8], inclusion problems, and fixed point problems [9–13]. Lemaire [12, 13] studied the relations between the well-posedness of minimization problems, inclusion problems, and fixed point problems. Fang et al. [11] proved that the well-posedness of a general mixed variational inequality is equivalent to the existence and the uniqueness of its solution in the Hilbert space. Recently, Ceng and Yao [9] got some results for the well-posedness of the generalized mixed variational inequality, the corresponding inclusion problem, and the corresponding fixed point problem. On the other hand, Li and Xia [14] considered the Levitin-Polyak well-posedness of a generalized variational inequality in Banach space. And they showed that the Levitin-Polyak well-posedness of a generalized variational inequality is equivalent to the uniqueness and existence of its solutions. However, there has been no result for the Levitin-Polyak well-posedness of an equilibrium-like problem.
Motivated and inspired by the research work going on in this field, in this paper, we extend the notion of Levitin-Polyak well-posedness to an equilibrium-like problem in Banach spaces and give some metric characterizations of its Levitin-Polyak well-posedness. Finally, we derive some conditions under which an equilibrium-like problem is Levitin-Polyak well-posed.
2. Preliminaries
Let X be a real reflexive Banach space with its dual X* and let K be a nonempty, closed, and convex subset of X. Let F:X→2X* be a set-valued mapping, and let ϕ:X*×X×X→R be a functional. In this paper, we consider the following equilibrium-like problem associated with (F,ϕ,K):
(1)ELP(F,ϕ,K):findx∈Ksuchthatforsomeu∈F(x),ϕ(u,x,y)≤0,∀y∈K.
Definition 1.
Let A,B be nonempty subsets of X. The Hausdorff metric H(·,·) between A and B is defined by
(2)H(A,B)=max{e(A,B),e(B,A)},
where e(A,B)=supa∈Ad(a,B) with d(a,B)=infb∈B∥a-b∥.
Lemma 2 (Nadler’s theorem [7]).
Let (X,∥·∥) be a normed vector space and let H(·,·) be the Hausdorff metric on the collection CB(X) of all nonempty, closed, and bounded subsets of X, induced by a metric d in terms of d(u,v)=∥u-v∥, which is defined by H(U,V)=max{e(U,V),e(V,U)}, for U and V in CB(X), where e(U,V)=supx∈Ud(x,V) with d(x,V)=infy∈V∥x-y∥. If U and V lie in CB(X), then, for any ϵ>0 and any u∈U, there exists v∈V such that ∥u-v∥≤(1+ϵ)H(U,V). In particular, whenever U and V are compact subsets in X, one has ∥u-v∥≤H(U,V).
Definition 3 (see [9]).
A nonempty set-valued mapping F:X→2X* is said to be
H-hemicontinuous if, for any x,y∈X, the function t↦H(F(x+t(y-x),F(x))) from [0,1] into R+=[0,+∞) is continuous at 0+, where H(·,·) is the Hausdorff metric defined on CB(X);
H-uniformly continuous if, for all ϵ>0, there exists δ>0 such that for all x,y∈X with ∥x-y∥<δ, one has H(F(x),F(y))<ϵ, where H(·,·) is the Hausdorff metric defined on CB(X).
Definition 4.
Let X and Y be two topological spaces and x∈X. A set-valued mapping F:X→2Y is said to be upper semicontinuous (u.s.c. in short) at x, if for any neighbourhood V of F(x), there exists a neighbourhood U of x such that F(y)⊂V, for all y∈U. If F is u.s.c. at each point of X, we say that F is u.s.c. on X.
Definition 5 (see [15]).
Let A be a nonempty subset of X. The measure of noncompactness μ of the set A is defined by
(3)μ(A)=inf{A⊂⋃i=1nAi,diamAi<ϵϵ>0:A⊂⋃i=1nAi,diamAi<ϵ,i=1,2,…,nA⊂⋃i=1nAi,diamAi<ϵ},
where diamAi denotes the diameter of the set Ai, for i=1,2,…,n.
Definition 6.
Let X be a real reflexive Banach space with its dual X* and let F:X→2X* be a set-valued mapping. A functional ϕ:X*×X×X→R is said to be monotone with respect to F, if for any x,y∈X and u∈F(x),v∈F(y), ϕ(u,x,y)≥ϕ(v,x,y).
Remark 7.
If ϕ(u,x,y)=〈u,x-y〉, for all x,y∈X and u∈F(x), it is easy to know that ϕ is monotone with respect to F which reduces to F being monotone.
We first prove the following proposition.
Proposition 8.
Let K be a nonempty, closed, and convex subset of X and let F:X→2X* be a nonempty compact-valued mapping which is H-hemicontinuous. Let ϕ:X*×X×X→R be monotone with respect to F, continuous in first argument, and concave in third argument. Moreover, ϕ(u,x,x)=0, for all u∈X*, x∈K. Then, for a given x∈K, the following statements are equivalent:
there exists u∈F(x) such that ϕ(u,x,y)≤0, for all y∈K;
ϕ(v,x,y)≤0, for all y∈K, v∈F(y).
Proof.
First, we assume that for some u∈F(x), ϕ(u,x,y)≤0, for all y∈K. Because ϕ is monotone with respect to F, we have
(4)ϕ(v,x,y)≤0,∀y∈K,v∈F(y).
Conversely, suppose that for all y∈K, v∈F(y), we obtain
(5)ϕ(v,x,y)≤0.
For any given y∈K, we define yt=ty+(1-t)x for all t∈(0,1). Replacing y by yt in the left-hand side of the last inequality, we have that, for each vt∈F(yt),
(6)0≥ϕ(vt,x,yt)=ϕ(vt,x,ty+(1-t)x)≥tϕ(vt,x,y)+(1-t)ϕ(vt,x,x)=tϕ(vt,x,y).
This implies that
(*)ϕ(vt,x,y)≤0,∀vt∈F(yt),t∈(0,1).
Since F:X→2X* is a nonempty compact-valued mapping, F(yt) and F(x) are nonempty compact and hence lie in CB(X). From Lemma 2, we get that, for each t∈(0,1) and for each fixed vt∈F(yt), there exists a ut∈F(x) such that
(7)∥vt-ut∥≤(1+t)H(F(yt),F(x)).
Since F(x) is compact, without loss of generality, we assume that ut→u∈F(x) as t→0+. Since F is H-hemicontinuous, we get that as t→0+,
(8)∥vt-ut∥≤(1+t)H(F(yt),F(x))⟶0.
This implies that vt→u∈F(x) as t→0+. Since ϕ is continuous in first argument, by (*) we obtain that there exists an u∈F(x) such that
(9)ϕ(u,x,y)≤0,∀y∈K.
This completes the proof.
3. Levitin-Polyak Well-Posedness of ELP(F,ϕ,K)
In this section, we extend the concepts of Levitin-Poylak well-posedness to the equilibrium-like problem and establish its metric characterizations. Let α≥0 be a given number, and let X, K, F, and ϕ be defined as the previous section.
Definition 9.
A sequence {xn}⊂X is called an LP α-approximating sequence for ELP(F,ϕ,K), if there exist wn∈X with wn→0 and 0<ϵn→0 such that xn+wn∈K for all n∈N and there exists un∈F(xn) such that
(10)ϕ(un,xn,y)≤α2∥xn-y∥2+ϵn,∀y∈K,n∈N.
If α1>α2≥0, then every LP α2-approximating sequence is LP α1-approximating. When α=0, we say that {xn} is an LP approximating sequence for ELP(F,ϕ,K).
Definition 10.
ELP(F,ϕ,K) is strongly LP α-well-posed if ELP(F,ϕ,K) has an unique solution and every LP α-approximating sequence converges strongly to the unique solution. In the sequel, strong LP 0-well-posedness is always called as strong LP well-posedness. If α1>α2≥0, then strong LP α1-well-posedness implies strong LP α2-well-posedness.
Definition 11.
ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense if ELP(F,ϕ,K) has nonempty solution set S and every LP α-approximating sequence has a subsequence which converges strongly to some point of S. In the sequel, strong LP 0-well-posedness in the generalized sense is always called as strong LP well-posedness in the generalized sense. If α1>α2≥0, then strong LP α1-well-posedness in the generalized sense implies strong LP α2-well-posedness in the generalized sense.
Remark 12.
If ϕ(u,x,y)=〈u,x-y〉+φ(x)-φ(y), for all x,y∈X, u∈F(x), then Definitions 10 and 11 reduce to Definitions 3.3 and 3.4 of [14], respectively. Moreover, when X is a Hilbert space, K=X, and wn≡0, Definitions 10 and 11 reduce to Definitions 3.2 and 3.3 of [11], respectively.
To obtain the metric characterizations of LP α-well-posedness, we consider the following LP α-approximating solution set of ELP(F,ϕ,K):
(11)Ωα(ϵ)={α2∥x-y∥2+ϵx∈domϕ:ffd(x,K)≤ϵ,ffandthereexistsu∈F(x)ffsuchthat∀y∈K,ϕ(u,x,y)≤α2∥x-y∥2+ϵ},fffff∀ϵ≥0.
Theorem 13.
Let K be a nonempty, closed, and convex subset of X and let F:X→2X* be a H-hemicontinuous and nonempty compact-valued mapping. Let ϕ:X*×X×X→R be monotone with respect to F, lower semicontinuous in second argument, and concave in third argument. Moreover, ϕ(u,x,x)=0, for all u∈X*, x∈K. Then, ELP(F,ϕ,K) is strongly LP α-well-posed if and only if
(12)Ωα(ϵ)≠∅,∀ϵ>0anddiam(Ωα(ϵ))⟶0asϵ⟶0.
Proof.
First, we assume that ELP(F,ϕ,K) is strongly LP α-well-posed and x*∈K is the unique solution of ELP(F,ϕ,K). It is easy to see that x*∈Ωα(ϵ). If diam(Ωα(ϵ))↛0 as ϵ→0, then there exist constant l>0 and sequences {ϵn}⊂R+ with ϵn→0 and {xn(1)},{xn(2)} with xn(1),xn(2)∈Ωα(ϵn) such that
(13)∥xn(1)-xn(2)∥>l,∀n∈N.
Because of xn(1),xn(2)∈Ωα(ϵn), by the definition of Ωα(ϵn), for xn(1), we obtain
(14)d(xn(1),K)≤ϵn<ϵn+1n,
and there exists un∈F(xn(1)) such that
(15)ϕ(un,xn(1),y)≤α2∥xn(1)-y∥2+ϵn,∀y∈K.
Since K is closed and convex, then there exists x¯n(1)∈K such that ∥xn(1)-x¯n(1)∥<ϵn+(1/n). Let wn=x¯n(1)-xn(1); we get wn+xn(1)=x¯n(1)∈K and ∥wn∥=∥xn(1)-x¯n(1)∥→0. This implies that wn→0. Thus, {xn(1)} is an LP approximating sequence for ELP(F,ϕ,K). By the similar argument, we obtain that {xn(2)} is an LP approximating sequence for ELP(F,ϕ,K). So they have to converge strongly to the unique solution of ELP(F,ϕ,K), which contradicts condition (13).
Conversely, suppose that condition (12) holds. Let {xn}⊂X be an LP α-approximating sequence for ELP(F,ϕ,K). Then, there exists wn∈X with wn→0 such that xn+wn∈K, and there exist 0<ϵn′→0 and un∈F(xn) such that
(16)ϕ(un,xn,y)≤α2∥xn-y∥2+ϵn′,∀y∈K,n∈N.
Since xn+wn∈K, then there exists x¯n∈K such that xn+wn=x¯n. It is obvious that d(xn,K)≤∥xn-x¯n∥=∥wn∥→0. Suppose that ϵn=max{ϵn′,∥wn∥}; we get that xn∈Ωα(ϵn). From (12), we have that {xn} is a Cauchy sequence and converges strongly to a point x¯∈K. Since ϕ is monotone with respect to F and lower semicontinuous in second argument, it follows from (16) that, for any y∈K, v∈F(y),
(17)ϕ(v,x¯,y)≤liminfn→∞{ϕ(v,xn,y)}≤liminfn→∞{ϕ(un,xn,y)}≤liminfn→∞{α2∥xn-y∥2+ϵn′}=α2∥x¯-y∥2.
For any y∈K, let yt=x¯+t(y-x¯), for all t∈[0,1]. Since K is a nonempty, closed, and convex subset, we have that yt∈K. Then, (17) implies that
(18)ϕ(vt,x¯,yt)≤α2∥x¯-yt∥2,∀vt∈F(yt).
Since ϕ is concave in third argument and ϕ(u,x,x)=0, for all u∈X*, x∈K,
(19)ϕ(vt,x¯,y)≤αt2∥x¯-y∥2,∀vt∈F(yt),y∈K.
Since F is a nonempty compact-valued mapping and H-hemicontinuous, by Lemma 2, for each fixed vt∈F(yt) and each t∈(0,1), there exists a ut∈F(x¯) such that ∥vt-ut∥≤H(F(yt),F(x¯)). Since F is H-hemicontinuous, we get that ∥vt-ut∥≤H(F(yt),F(x¯))→0 as t→0+. Since F is compact, without loss of generality, we assume that ut→u∈F(x¯) as t→0+. Thus, we obtain that
(20)∥vt-u∥≤∥vt-ut∥+∥ut-u∥≤H(F(yt),F(x¯))+∥ut-u∥⟶0ast⟶0+.
This implies that vt→u as t→0+. It follows from (19) that
(21)ϕ(u,x¯,y)≤0,∀y∈K.
Therefore, x¯ solves ELP(F,ϕ,K).
To complete the proof, we only need to prove that ELP(F,ϕ,K) has a unique solution. Suppose that ELP(F,ϕ,K) has two distinct solutions x1 and x2. Then, it is obvious that x1,x2∈Ωα(ϵ) for all ϵ>0 and
(22)0<∥x1-x2∥≤diam(Ωα(ϵ))⟶0,
a contradiction to (12). This completes the proof.
Theorem 14.
Let K be a nonempty, closed, and convex subset of X and let F:X→2X* be a H-hemicontinuous and nonempty compact-valued mapping. Let ϕ:X*×X×X→R be monotone with respect to F and lower semicontinuous in second argument. Moreover, ϕ(u,x,x)=0, for all u∈X*, x∈K. Then, ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense if and only if
(23)Ωα(ϵ)≠∅,∀ϵ>0andμ(Ωα(ϵ))⟶0asϵ⟶0.
Proof.
Assume that ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense. Let S be the solution set of ELP(F,ϕ,K). Then, S is nonempty and compact. Indeed, let {xn} be any sequence in S. Then, {xn} is an LP α-approximating sequence for ELP(F,ϕ,K). Since ELP(F,ϕ,K) is strongly α-well-posed in the generalized sense, {xn} has a subsequence which converges strongly to some point of S. Thus, S is compact. It is easy to see that Ωα(ϵ)⊃S≠∅ for all ϵ>0. Now we show that
(24)μ(Ωα(ϵ))⟶0asϵ⟶0.
It is easy to see that, for every ϵ>0,
(25)H(Ωα(ϵ),S)=max{e(Ωα(ϵ),S),e(S,Ωα(ϵ))}=e(Ωα(ϵ),S).
Taking into account the compactness of S, we obtain
(26)μ(Ωα(ϵ))≤2H(Ωα(ϵ),S)+μ(S)=2e(Ωα(ϵ),S).
To prove (23), it is sufficient to show that
(27)e(Ωα(ϵ),S)⟶0asϵ⟶0.
Indeed, if e(Ωα(ϵ),S)↛0 as ϵ→0, then there exist l>0 and {ϵn}⊂R+ with ϵn→0, and xn∈Ωα(ϵn) such that
(28)xn∉S+B(0,l),∀n∈N,
where B(0,l) is the closed ball centered at 0 with radius l. By the definition of Ωα(ϵn), we know that d(xn,K)≤ϵn<ϵn+(1/n), and there exists un∈F(xn) such that
(29)ϕ(un,xn,y)≤α2∥xn-y∥2+ϵn,∀y∈K.
Thus, there exists x¯n∈K such that ∥x¯n-xn∥<ϵn+(1/n). Let wn=x¯n-xn; then, we have wn+xn∈K with wn→0. So {xn} is an LP α-approximating sequence for ELP(F,ϕ,K). Since ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense, there exists a subsequence {xnk} of {xn} which converges strongly to some point of S. This contradicts (28) and so
(30)e(Ωα(ϵ),S)⟶0asϵ⟶0.
Conversely, suppose that (23) holds. We first show that Ωα(ϵ) is closed for all ϵ>0. Let {xn}⊂Ωα(ϵ) with xn→x; then, there exists un∈F(xn) such that d(xn,K)≤ϵ and
(31)ϕ(un,xn,y)≤α2∥xn-y∥2+ϵ,∀y∈K,n∈N.
Since F is an upper semicontinuous and nonempty compact-valued mapping, there exist a sequence {unk} of {un} and some u∈F(x) such that unk→u. Therefore, it follows from (31) and the lower semicontinuity of ϕ that
(32)ϕ(u,x,y)≤α2∥x-y∥2+ϵ,∀y∈K.
It is obvious that d(x,K)≤ϵ. This implies that x∈Ωα(ϵ) and so Ωα(ϵ) is nonempty closed for all ϵ>0. Observe that
(33)S=⋂ϵ>0Ωα(ϵ).
Since μ(Ωα(ϵ))→0, the theorem in page 412 of [15] can be applied and one concludes that S is nonempty and compact with
(34)e(Ωα(ϵ),S)=H(Ωα(ϵ),S)⟶0.
Let {x^n}⊂X be an LP α-approximating sequence for ELP(F,ϕ,K). Then, there exists wn∈X with wn→0 such that x^n+wn∈K, and there exist u^n∈F(x^n) and 0<ϵn′→0 such that
(35)ϕ(u^n,x^n,y)≤α2∥x^n-y∥2+ϵn′,∀y∈K,n∈N.
Since x^n+wn∈K, then there exists x¯n∈K such that x^n+wn=x¯n. It follows that
(36)d(x^n,K)≤∥x^n-x¯n∥=∥wn∥⟶0.
Set ϵn=max{∥wn∥,ϵn′}; we get x^n∈Ωα(ϵn). From (23) and the definition of Ωα(ϵn), we obtain
(37)d(x^n,S)≤e(Ωα(ϵn),S)⟶0.
Since S is compact, there exists pn∈S such that
(38)∥pn-x^n∥=d(x^n,S)⟶0.
From the compactness of S, there exists a subsequence {pnk} of {pn} which converges strongly to p¯∈S. Hence, the corresponding subsequence {x^nk} of {x^n} converges strongly to p¯∈S. Thus, ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense. The proof is complete.
4. Conditions for Levitin-Polyak Well-Posedness
In this section, we get some conditions under which the ELP(F,ϕ,K) in Banach spaces is Levitin-Polyak well-posed.
For any δ0≥0, we denote M(δ0)={x∈X:dK(x)≤δ0}. We have the following result.
Theorem 15.
Let K be a nonempty, closed, and convex subset of X and let F:X→2X* be a H-hemicontinuous and nonempty compact-valued mapping. Let ϕ:X*×X×X→R be monotone with respect to F, lower semicontinuous in first and second arguments, and concave in third argument. Moreover, ϕ(u,x,x)=0, for all u∈X*, x∈K. If there exists some δ0 with δ0>0 such that M(δ0) is compact, then ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense.
Proof.
Let {xn} be an LP approximating sequence for ELP(F,ϕ,K). Then, there exist 0<ϵn′→0 and wn∈X with wn→0 such that
(39)xn+wn∈K,
and there exists un∈F(xn) satisfying
(40)ϕ(un,xn,y)≤α2∥xn-y∥2+ϵn′,∀y∈K,n∈N.
Since xn+wn∈K, then there exists x¯n∈K such that xn+wn=x¯n. Thus,
(41)d(xn,K)≤∥xn-x¯n∥=∥wn∥⟶0.
Let ϵn=max{ϵn′,∥wn∥}; we can get d(xn,K)≤ϵn. Without loss of generality, suppose that {xn}⊂M(δ0) for n is sufficiently large. By the compactness of M(δ0), there exist a subsequence {xnk} of {xn} and x¯∈M(δ0) such that xnk→x¯. It is easy to see that x¯∈K. Furthermore, by the u.s.c. of F at x¯ and compactness of F(x¯), there exist a subsequence {unk} of {un} and some u¯∈F(x¯) such that unk→u¯. Since ϕ is lower semicontinuous in first and second arguments, it follows from (40) that
(42)ϕ(u¯,x¯,y)≤α2∥x¯-y∥2,∀y∈K.
For any y∈K, let yt=x¯+t(y-x¯), for all t∈(0,1); it is obvious that yt∈K. Now, from (42), we have
(43)ϕ(u¯,x¯,yt)≤α2∥x¯-yt∥2.
By the convexity of ϕ, it follows that, for each t∈(0,1), we obtain
(44)ϕ(u¯,x¯,y)≤αt2∥x¯-y∥2,∀y∈K.
Let t→0+ in the last inequality; then, we have
(45)ϕ(u¯,x¯,y)≤0,∀y∈K.
This shows that x¯ solves ELP(F,ϕ,K). Thus, ELP(F,ϕ,K) is strongly LP α-well-posed in the generalized sense.
Conflict of Interests
The author declares that there is no conflict of interests regarding the publication of this paper.
TykhonovA. N.On the stability of the functional optimization problem19666631634ZolezziT.Extended well-posedness of optimization problems19969112572662-s2.0-004000821110.1007/BF02192292ZBL0873.90094ZolezziT.Well-posedness criteria in optimization with application to the calculus of variations19952554374532-s2.0-000116170410.1016/0362-546X(94)00142-5ZBL0841.49005ZolezziT.Well-posedness of optimal control problems199423289301ZBL0810.49010LevitinE. S.PolyakB. T.Convergence of minimizing sequences in conditional extremum problem19967764767BednarczukE. M.JahnJ.KrabsW.Well-posedness of optimization problem1987294Berlin, GermanySpringer5161Lecture Notes in Economics and Mathematical Systems10.1007/978-3-642-46618-2_2LucchettiR.PatroneF.A characterization of Tykhonov well-posedness for minimum problems with applications to variational inequalities1981344614762-s2.0-001966699110.1080/01630568108816100LucchettiR.PatroneF.Hadamard and Tyhonov well-posedness of a certain class of convex functions19828812042152-s2.0-001131008210.1016/0022-247X(82)90187-1ZBL0487.49013CengL. C.YaoJ. C.Well-posedness of generalized mixed variational inequalities, inclusion problems and fixed-point problems20086912458546032-s2.0-5504910245810.1016/j.na.2007.11.015ZBL1153.49024FangY.-P.HuangN.-J.YaoJ.-C.Well-posedness by perturbations of mixed variational inequalities in Banach spaces201020136826922-s2.0-7034946469610.1016/j.ejor.2009.04.001ZBL1177.49018FangY.-P.HuangN.-J.YaoJ.-C.Well-posedness of mixed variational inequalities, inclusion problems and fixed point problems20084111171332-s2.0-4154910886010.1007/s10898-007-9169-6ZBL1149.49009LemaireB.conditions, and regularization of minimization, inclusion, and fixed-point problems1998127184LemaireB.Ould Ahmed SalemC.RevalskiJ. P.Well-posedness by perturbations of variational problems200211523453682-s2.0-003641508210.1023/A:1020840322436ZBL1047.90067LiX.-B.XiaF.-Q.Levitin-Polyak well-posedness of a generalized mixed variational inequality in Banach spaces2012754213921532-s2.0-8465516803110.1016/j.na.2011.10.013ZBL1237.49016KuratowskiK.19681-2New York, NY, USAAcademic Press