Jump to ContentJump to Main Navigation
Show Summary Details
More options …

Open Mathematics

formerly Central European Journal of Mathematics

Editor-in-Chief: Gianazza, Ugo / Vespri, Vincenzo


IMPACT FACTOR 2018: 0.726
5-year IMPACT FACTOR: 0.869

CiteScore 2018: 0.90

SCImago Journal Rank (SJR) 2018: 0.323
Source Normalized Impact per Paper (SNIP) 2018: 0.821

Mathematical Citation Quotient (MCQ) 2018: 0.34

ICV 2017: 161.82

Open Access
Online
ISSN
2391-5455
See all formats and pricing
More options …
Volume 17, Issue 1

Issues

Volume 13 (2015)

A categorical approach to abstract convex spaces and interval spaces

Bing Wang / Qing-Hua Li / Zhen-Yu Xiu
  • College of Applied Mathematics, Chengdu University of Information Technology, Chengdu, 610000, China
  • College of Mathematics and Econometrics, Hunan University, Hunan, 410082, China
  • Email
  • Other articles by this author:
  • De Gruyter OnlineGoogle Scholar
Published Online: 2019-05-16 | DOI: https://doi.org/10.1515/math-2019-0029

Abstract

In this paper, we establish the axiomatic conditions of hull operators and introduce the category of interval spaces. We also investigate their relations with convex spaces from a categorical sense. It is shown that the category CS of convex spaces is isomorphic to the category HS of hull spaces, and they are all topological over Set. Also, it is proved that there is an adjunction between the category IS of interval spaces and the category CS of convex spaces. In particular, the category CS(2) of arity 2 convex spaces can be embedded in IS as a reflective subcategory.

Keywords: Convex structure; Hull operator; Interval operator; Adjunction

MSC 2010: 52A01; 54A05; 18D35

1 Introduction

Convexity is an important and basic property in many mathematical areas. However, in some concrete mathematical setting, such as vector spaces, it is not the most suitable setting for studying the basic properties of convex sets. In order to avoid this deficiency, abstract convex structures (convex structures, in short) are defined by three axioms [1], which is a similar way of defining topological structures. Up to now, the convexity theory has become a branch of mathematics dealing with set-theoretic structures satisfying axioms similar to that usual convex sets fulfill. Actually, convex structures appeared in many research areas, such as lattices [2], graphs [3], and topology [4]. Besides, convexity theory is also investigated from the lattice-valued aspect, including L-convex structures [5, 6, 7, 8, 9, 10, 11, 12, 13] and M-fuzzifying convex structures [14, 15, 16, 17, 18, 19].

Category theory plays an important role in demonstrating the relations between different types of spatial structures. It emerges frequently in general topology and fuzzy topology, especially in crisp and fuzzy convergence theory [20, 21, 22, 23, 24, 25, 26, 27, 28, 29]. This motivates us to apply category theory to convex structures since convex structures can be considered as topology-like structures. Actually, like continuous mappings between topological spaces, there is also a special kind of mappings between convex spaces, which is called convexity-preserving mapping. Under a convexity-preserving mapping, convex sets in the range space are inverted to convex sets of the domain. Such mappings arise in various constructions of convexities. For spaces derived from an algebraic structure, convexity-preserving mappings usually agree with the corresponding notions of homomorphisms. In fact, convexity-preserving mappings are exactly the appropriate notions of morphisms in the category of convex spaces. This is just one of our motivations of this paper. That is, we would like to investigate the categorical properties of convex spaces.

A convex structure is completely determined by a special kind of closure operators, which is called algebraic closure operators. Actually, algebraic closure operators can be treated as the hull operators of convex spaces. Except for algebraic closure operators, interval operators, as a generalization of intervals, provide a natural and frequent method of describing or constructing convex structures. There are also close relations between convex structures and interval operators. Inspired by this, we will not only provide a new characterization of convex structures by closure operators and present an axiomatic hull operators, but also focus on the categorical properties of interval spaces and study its relations with convex spaces from a categorical sense.

2 Preliminaries

Throughout this paper, let X denote a nonempty set and 2X the powerset of X. Let {Aj}jJ dir 2X denote that {Aj}jJ is a directed subset of 2X, which means for each B, C ∈ {Aj}jJ, there exists D ∈ {Aj}jJ such that BD and CD.

Let X, Y be two nonempty sets and f : XY be a mapping. Define f : 2X ⟶ 2Y and f : 2Y ⟶ 2X as follows:

A2X,f(A)={f(x)xA};B2Y,f(B)={xf(x)B}.

Definition 2.1

([1]). A convex structure 𝓒 on X is a subset of 2X which satisfies:

  • (CS1)

    ∅, X ∈ 𝓒;

  • (CS2)

    {Ai}iI ⊆ 𝓒 implies ⋂iI Ai ∈ 𝓒;

  • (CS3)

    {Aj}jJ dir 𝓒 implies dirjJC.

    For a convex structure 𝓒 on X, the pair (X, 𝓒) is called a convex space.

A mapping f : (X, 𝓒X) ⟶ (Y, 𝓒Y) is called convexity-preserving (CP, in short) provided that B ∈ 𝓒Y implies f(B) ∈ 𝓒X. The category whose objects are convex spaces and whose morphisms are CP mappings will be denoted by CS.

Definition 2.2

([1]). A closure operator on X is a mapping C : 2X ⟶ 2X which satisfies:

  • (CL1)

    C(∅) = ∅;

  • (CL2)

    AC(A);

  • (CL3)

    ABC(A) ⊆ C(B);

  • (CL4)

    CC(A) = C(A).

For a closure operator C on X, the pair (X, C) is called a closure space. It will be called algebraic if it also satisfies that,

(CLA) C(A) = ⋃{C(B) | B is a finite subset of A}.

The pair (X, C) is called an algebraic closure space.

A mapping f : (X, CX) ⟶ (Y, CY) between closure spaces is called convexity-preserving (CP, in short) provided that f(CX(A)) ⊆ CY(f(A)) for all A ∈ 2X. The category whose objects are closure spaces and whose morphisms are CP mappings will be denoted by CLS, and the full subcategory of algebraic closure spaces by ACLS.

For a convex space (X, 𝓒), define C𝓒 : 2X ⟶ 2X by C𝓒(A) = ⋂AB∈𝓒 B. Then C𝓒 is an algebraic closure operator on X. Conversely, for a closure operator (X, C), define 𝓒C ⊆ 2X by 𝓒C = {A ∈ 2X | A = C(A)}. Then 𝓒C is a convex structure on X. Furthermore, they are one-to-one corresponding. In a categorical sense, we have

Theorem 2.3

The category CS is isomorphic to the category ACLS.

Definition 2.4

([29, 30]). A category C is called a topological category over Set with respect to the usual forgetful functor from C to Set if it satisfies (TC1), (TC2) and (TC3) or (TC1), (TC2) and (TC3).

(TC1) Existence of initial structures: For any set X, any class J, and family ((Xj, ξj))jJ of C-object and any family (fj : XXj)jJ of mappings, there exists a unique C-structure ξ on X which is initial with respect to the source (fj : X ⟶ (Xj, ξj))jJ, this means that for a C-object (Y, η), a mapping g : (Y, η) ⟶ (X, ξ) is a C-morphism iff for all jJ, fjg :(Y, η) ⟶ (Xj, ξj) is a C-morphism.

(TC1) Existence of final structures: For any set X, any class J, and family ((Xj, ξj))jJ of C-object and any family (fj : XjX)jJ of mappings, there exists a unique C-structure ξ on X which is final with respect to the sink (fj : (Xj, ξj) ⟶ X))jJ, this means that for a C-object (Y, η), a mapping g : (X, ξ) ⟶ (Y, η) is a C-morphism iff for all jJ, gfj : (Xj, ξj) ⟶ (Y, η) is a C-morphism.

(TC2) Fibre-smallness: For any set X, the C-fibre of X, i.e., the class of all C-structures on X, which we denote by C(X), is a set.

(TC3) Terminal separator property: For any set X with cardinality at most one, there exists exactly one C-object with underlying set X (i.e. there exists exactly one C-structure on X).

Lemma 2.5

([29, 30]). Suppose that 𝔽 : AB and 𝔾 : BA are concrete functors. Then the following conclusions are equivalent:

  1. {idY : 𝔽 ∘ 𝔾(Y) ⟶ Y | YB} is a natural transformation from the functor 𝔽 ∘ 𝔾 to the identity functor idB on B, and {idX : X ⟶ 𝔾 ∘ 𝔽(X) | XA} is a natural transformation from the identity functor idA on A to the functor 𝔾 ∘ 𝔽.

  2. For each YB, idY : 𝔽 ∘ 𝔾(Y) ⟶ Y is a B-morphism, and for each XA, idX : X ⟶ 𝔾 ∘ 𝔽(X) is a A-morphism.

    In this case, (𝔽, 𝔾) is called an adjunction between A and B.

If (𝔽, 𝔾) is an adjunction, then it is easy to verify that 𝔽 is a left adjoint of 𝔾 or equivalently, 𝔾 is a right adjoint of 𝔽.

The class of objects of a category A is denoted by |A|. For more notions related to category theory we refer to [29] and [30].

3 The categories of convex spaces and hull spaces

In this section, we first study some categorical properties of convex spaces. Then we introduce the axiomatic hull operators and investigate its relations with convex structures.

Definition 3.1

For a nonempty, let F𝓒(X) denote the fibre

{(X,C)Cis a convex structure on X.}

of X. For convex spaces (X, 𝓒1) and (X, 𝓒2), we say (X, 𝓒1) is finer than (X, 𝓒2), or (X, 𝓒2) is coarser than (X, 𝓒1), denoted by (X, 𝓒1) ⩽𝓒 (X, 𝓒2), if the identity mapping idX : (X, 𝓒1) ⟶ (X, 𝓒2) is CP. We also write 𝓒1𝓒 𝓒2.

Example 3.2

Let X be a nonempty set.

  1. Define 𝓒* by 𝓒* = 2X. Then 𝓒* is the finest convex structure on X, which is called the discrete convex structure on X.

  2. Define 𝓒* by 𝓒* = {∅, X}. Then 𝓒* is the coarsest convex structure on X, which is called the indiscrete convex structure on X.

Theorem 3.3

The category CS is topological over Set.

Proof

We first prove the existence of final structures. Let ((Xλ, 𝓒λ))λΛ be a family of convex spaces and let X be a nonempty set. Let further ((fλ : (Xλ, 𝓒λ) ⟶ X))λΛ be a sink. Define 𝓒 ⊆ 2X by

C={A2XλΛ,fλ(A)Cλ}.

Since fλ preserves arbitrary meets and directed joins, we can verify that 𝓒 is a convex structure on X.

Let further (Y, 𝓒Y) be a convex space and g : XY be a mapping. Assume that gfλ is CP for all λΛ. We have for all B ∈ 𝓒Y,

λΛ,fλ(g(B))=(gfλ)(B)Cλ.

By definition of 𝓒, we obtain g(B) ∈ 𝓒. This implies that g : (X, 𝓒) ⟶ (Y, 𝓒Y) is CP, as desired.

Secondly, the class of all convex structures on a fixed set X is a subset of 2(2X), which means that the CS fibre of X is a set.

Finally, for a one point set X = {x}, there exists only one convex structure 𝓒 = {∅, {x}} on X. Hence, CS satisfies the terminal separator property. Therefore, CS is a topological category in the sense of [29]. That is, a well-fibred topological category in the terminology of [30].□

Corollary 3.4

(F𝓒(X), ⩽𝓒) is a complete lattice.

Next we will introduce the axiomatic hull operators and study its relations with convex structures.

Definition 3.5

A hull operator on X is a mapping co : 2X ⟶ 2X which satisfies:

  • (H1)

    co(∅) = ∅, co(X) = X;

  • (H2)

    Aco(A);

  • (H3)

    ABco(A)⊆ co(B);

  • (H4)

    co(co(A)) = co(A).

  • (H5)

    co(dirjJAj) = ⋃jJ co(Aj).

For a hull operator co on X, the pair (X, co) is called a hull space. Actually, a hull operator on X is a closure operator on X which satisfies (H5).

Definition 3.6

A mapping f : (X, coX) ⟶ (Y, coY) is called convexity-preserving (CP, in short) provided that f(coX(A)) ⊆ coY(f(A)) for all A ∈ 2X.

The category whose objects are hull spaces and whose morphisms are CP mappings will be denoted by HS.

Proposition 3.7

Let (X, 𝓒) be a convex space and define co𝓒 : 2X ⟶ 2X by

A2X,coC(A)=ABCB.

Then co𝓒 is a hull operator on X.

Proof

By Theorem 2.3, co𝓒 is an algebraic closure operator on X. Thus co𝓒 satisfies (H1)–(H4). It suffices to verify (H5). For {Aj}jJ dir 2X, take any x ∉ ⋃jJ co𝓒(Aj) = ⋃jJAjB∈𝓒 B. Then there exists Bj ∈ 2X such that AjB ∈ 𝓒 and xBj for each jJ. Let Cj = co𝓒(Aj). By (C2) and (H3), we know AjCj ∈ 𝓒 and {Cj}jJ is directed. Put B = ⋃jJ Cj. By (C3), we obtain ⋃jJ AjB ∈ 𝓒. Further, since CjBj, it follows that xCj for each jJ. This implies that xB. As a consequence, we obtain B ∈ 2X such that ⋃jJ AjB ∈ 𝓒 and xB. This means that xco𝓒(dirjJ Aj). By the arbitrariness of x, we have co𝓒(dirjJ Aj) ⊆ ⋃jJ co𝓒(Aj). The inverse inequality holds obviously. Therefore, co𝓒 is a hull operator.□

Proposition 3.8

If f : (X, 𝓒X) ⟶ (Y, 𝓒Y) is a CP mapping, then so is f : (X, co𝓒X) ⟶ (Y, co𝓒Y).

Proof

Since f : (X, 𝓒X) ⟶ (Y, 𝓒Y) is a CP mapping, it follows that

B2Y,BCYimpliesf(B)CX.

Then for each A ∈ 2X, we have

f(coCY(f(A)))=f(A)BCYf(B)Af(B)CXf(B)ACCXC=coCX(A).

This shows f(co𝓒X(A)) ⊆ co𝓒Y(f(A)), as desired.□

By Propositions 3.7 and 3.8, we obtain a functor 𝔽 : CSHS as follows:

F:CSHS(X,C)(X,coC)ff.

Proposition 3.9

Let (X, co) be a hull space and define 𝓒co = {A ∈ 2X | A = co(A)}. Then 𝓒co is a convex structure on X.

Proof

(C1) is obvious. We need only verify (C2) and (C3).

(C2) Take any {Ai}iI ⊆ 𝓒co. Then for each iI, co(Ai) = Ai. By (H3), co(⋂iI Ai) ⊆ ⋂iI co(Ai). In order to show the inverse inequality, take any xco(⋂iI Ai) ⊇ ⋂iI Ai. Then there exists i0I such that xAi0 = co(Ai0). This implies that x ∉ ⋂iI co(Ai). By the arbitrariness of x, we obtain ⋂iI co(Ai) ⊆ co(⋂iI Ai). Hence, it follows that ⋂iI co(Ai) = co(⋂iI Ai). This means that ⋂iI Ai ∈ 𝓒co.

(C3) Take any {Aj}jJ dir 𝓒co. Then Aj = co(Aj) for each jJ. By (H5), it follows that

co(dirjJAj)=dirjJco(Aj)=dirjJAj.

This means that dirjJ Aj ∈ 𝓒co.□

Proposition 3.10

If f : (X, coX) ⟶ (Y, coY) is a CP mapping, then so is f : (X, 𝓒coX) ⟶ (Y, 𝓒coY).

Proof

Since f : (X, coX) ⟶ (Y, coY) is a CP mapping, we have

A2X,coX(A)f(coY(f(A))).

Then for each B ∈ 𝓒coY, it follows that

coX(f(B))f(coY(f(f(B))))f(coY(B))=f(B).

This implies that coX(f(B)) = f(B). Hence, f(B) ∈ 𝓒coX.□

By Propositions 3.9 and 3.10, we obtain a functor 𝔾 : HSCS as follows:

G:HSCS(X,co)(X,Cco)ff.

Theorem 3.11

The category CS is isomorphic to the category HS.

Proof

It suffices to show that 𝔾 ∘ 𝔽 = 𝕀CS and 𝔽 ∘ 𝔾 = 𝕀HS. That is, for each (X, 𝓒) ∈ |CS| and each (X, co) ∈ |HS|, co𝓒co = co and 𝓒co𝓒 = 𝓒.

For each A ∈ 2X, we have

coCco(A)=ABCcoB=AB=co(B)B=co(A)

and

ACcoCA=coC(A)=ABCBAC.

This completes the proof.□

4 The category of interval spaces

In this section, we will introduce the category of interval spaces with interval spaces as objects and with interval-preserving mappings as morphisms. Then we will study its categorical properties.

Definition 4.1

([1]). An interval operator on X is a mapping 𝓘 : X × X ⟶ 2X which satisfies:

  • (I1)

    x, y ∈ 𝓘(x, y);

  • (I2)

    𝓘(x, y) = 𝓘(y, x).

For an interval operator 𝓘 on X, the pair (X, 𝓘) is called an interval space.

Definition 4.2

([1]). A mapping f : (X, 𝓘X) ⟶ (Y, 𝓘Y) is called interval-preserving (IP in short) provided that

x,yX,f(JX(x,y))JY(f(x),f(y)).

It is easy to check that all interval spaces and IP mappings form a category, denoted by IS.

Definition 4.3

For a nonempty set X, let F𝓘(X) denote the fibre

{(X,J)Jis an interval operator on X}

of X. For interval spaces (X, 𝓘1) and (X, 𝓘2), we say (X, 𝓘1) is finer than (X, 𝓘2), or (X, 𝓘2) is coarser than (X, 𝓘1), denoted by (X, 𝓘1) ⩽𝓘 (X, 𝓘2), if the identity mapping idX : (X, 𝓘1) ⟶ (X, 𝓘2) is IP. We also write 𝓘1𝓘 𝓘2.

Example 4.4

Let X be a nonempty set.

  1. Define 𝓘* : X × X ⟶ 2X by 𝓘*(x, y) = {x, y} for each x, yX. Then 𝓘* is the finest interval operator on X, which is called the discrete interval operator on X.

  2. Define 𝓘* : X × X ⟶ 2X by 𝓘*(x, y) = X. Then 𝓘* is the coarsest interval operator on X, which is called the indiscrete interval operator on X.

  3. Suppose that ℝ is the set of real numbers. Define 𝓘 : ℝ × ℝ ⟶ 2 by

    a,bR,JR(a,b)=[min{a,b},max{a,b}].

    Then 𝓘 is an interval operator on ℝ.

  4. Suppose that d is a metric on X. Define 𝓘d : X × X ⟶ 2X by

    x,yX,Jd(x,y)={zXd(x,y)=d(x,z)+d(z,y)}.

Then 𝓘d is an interval operator on X.

Proposition 4.5

Let (X, 𝓘X), (Y, 𝓘Y) and (Z, 𝓘Z) be interval spaces. If f : XY and g : YZ are IP, then gf : (X, 𝓘X) ⟶ (Z, 𝓘Z) is IP.

Proof

The proof is easy and omitted.□

In the category IS, important constructions like the formulations of products and subspaces, are always possible.

Theorem 4.6

The category IS is topological over Set.

Proof

We first prove the existence of initial structures. Let ((Xλ, 𝓘λ))λΛ be a family of interval spaces and let X be a nonempty set. Let further ((fλ : X ⟶ (Xλ, 𝓘λ)))λΛ be a source. Define 𝓘 : X × X ⟶ 2X by

x,yX,J(x,y)=λΛfλ(Jλ(fλ(x),fλ(y))).

Then it is easy to verify that 𝓘 is an interval operator on X.

Let further (Y, 𝓘Y) be an interval space and g : YX be a mapping. Assume that fλg is IP for all λΛ. We then have for each y1, y2Y and for each λΛ,

(fλg)(JY(y1,y2))Jλ(fλ(g(y1)),fλ(g(y2))).

From this we obtain

g(JY(y1,y2))λΛfλ(Jλ(fλ(g(y1)),fλ(g(y2))))=J(g(y1),g(y2)).

This means that g : (Y, 𝓘Y) ⟶ (X, 𝓘) is IP, as desired.

Secondly, the class of all interval operators on a fixed set X is a subset of 2((2X)X×X), which means that the IS fibre of X is a set.

Finally, for a one point set X = {x}, there exists only one interval operator 𝓘 on X, which is defined by 𝓘(x, x) = X. Hence, IS satisfies the terminal separator property. Therefore, IS is a topological category over Set.□

Corollary 4.7

(F𝓘(X), ⩽𝓘) is a complete lattice.

Example 4.8

(Product Spaces). Let {(Xλ, 𝓘λ)}λΛ be a family of interval spaces. The interval operator Π – 𝓘 on ∏λΛ Xλ which is initial with respect to the projections (pλ)λΛ is called the product interval operator and the pair (∏λΛ Xλ, Π – 𝓘) is called the product space. By definition, we have for x, y ∈ ∏λΛ Xλ,

ΠJ(x,y)=λΛpλ(Jλ(pλ(x),pλ(y)))=λΛJλ(pλ(x),pλ(y)).

Example 4.9

(Subspaces) Let (X, 𝓘X) be an interval space and let YX. The interval operator 𝓘Y on Y which is initial with respect to the inclusion mapping idY : YX is called the sub-interval operator and the pair (Y, 𝓘Y) is called the subspace of (X, 𝓘X). By definition, we have for x, yY,

JY(x,y)=JX(x,y)Y.

5 Relations between IS and CS

In this section, we will focus on the relations between IS and CS. In particular, we will propose a full subcategory of CS, consisted of arity 2 convex spaces and study its relations with interval spaces.

Definition 5.1

A convex space (X, 𝓒) is called arity 2 if it satisfies

(AR2) ∀A ∈ 2X, ∀x, yA, co𝓒({x, y}) ⊆ A implies A ∈ 𝓒.

Let CS(2) denote the full subcategory of CS, consisted of arity 2 convex spaces.

Next we will study the relations between CS (CS(2)) and IS.

Proposition 5.2

Let (X, 𝓒) be a convex space and define 𝓘𝓒 : X × X ⟶ 2X by

x,yX,JC(x,y)=coC({x,y})=x,yACA.

Then 𝓘𝓒 is an interval operator on X.

Proof

The verifications of (I1) and (I2) are straightforward and omitted.□

Proposition 5.3

If f : (X, 𝓒X) ⟶ (Y, 𝓒Y) is a CP mapping, then f : (X, 𝓘𝓒X) ⟶ (Y, 𝓘𝓒Y) is a IP mapping.

Proof

Since f : (X, 𝓒X) ⟶ (Y, 𝓒Y) is a CP mapping, it follows that f(B) ∈ 𝓒X for each B ∈ 𝓒Y. Then for each x, yX, we have

f(JCY(f(x),f(y)))=f(x),f(y)BCYf(B)x,yf(B)CXf(B)x,yACXA=JCX(x,y).

This implies that f(𝓘𝓒X(x, y)) ⊆ 𝓘𝓒Y(f(x), f(y)), as desired.□

By Propositions 5.2 and 5.3, we obtain a functor ℍ as follows:

H:CSIS(X,C)(X,JC)ff.

Proposition 5.4

Let (X, 𝓘) be an interval space and define 𝓒𝓘 as follows:

CJ={A2Xx,yA,J(x,y)A}.

Then (X, 𝓒𝓘) is an arity 2 convex space.

Proof

(C1) is obvious. We need only verify (C2), (C3) and (AR2).

(C2) Take any {Ai}iI ⊆ 𝓒𝓘. Then for each iI and for each x, yAi, 𝓘(x, y) ⊆ Ai. This implies that

x,yiIAiiI,x,yAiiI,J(x,y)AiJ(x,y)iIAi.

Hence, ⋂iI Ai ∈ 𝓒𝓘.

(C3) Take any {Aj}jJ dir 𝓒𝓘. Then for each x, ydirjJ Aj, there exist j1, j2J such that xAj1 and yAj2. Since {Aj}jJ is directed, there exists j3J such that Aj1Aj3 and Aj2Aj3. This implies that x, yAj3 ∈ 𝓒𝓘. Then it follows that 𝓘(x, y) ⊆ Aj3dirjJ Aj. This means dirjJ Aj ∈ 𝓒𝓘.

(AR2) Take any A ∈ 2X such that

x,yA,coCJ({x,y})A.

In order to show A ∈ 𝓒𝓘, take any x, yA. It follows that

coCJ({x,y})=x,yBCJBJ(x,y).

Then we have 𝓘(x, y) ⊆ co𝓒𝓘({x, y}) ⊆ A for each x, yA. This implies that A ∈ 𝓒𝓘, as desired.□

Proposition 5.5

If f : (X, 𝓘X) ⟶ (Y, 𝓘Y) is a IP mapping, then f : (X, 𝓒𝓘X) ⟶ (Y, 𝓒𝓘Y) is a CP mapping.

Proof

Since f : (X, 𝓘X) ⟶ (Y, 𝓘Y) is a IP mapping, it follows that

x,yX,f(JX(x,y))JY(f(x),f(y)).

Then for each B ∈ 𝓒𝓘Y, take any x, yf(B). It follows that f(x), f(y) ∈ B ∈ 𝓒𝓘Y. This means that 𝓘Y(f(x), f(y)) ⊆ B. Further we have

JX(x,y)f(JY(f(x),f(y)))f(B).

This implies that f(B) ∈ 𝓒𝓘X, as desired.□

By Propositions 5.4 and 5.5, we obtain a functor 𝕂 : ISCS as follows:

K:ISCS(X,J)(X,CJ)ff.

Theorem 5.6

(𝕂, ℍ) is an adjunction between IS and CS.

Proof

Since 𝕂 and ℍ are both concrete functors, we need only verify that 𝕂 ∘ ℍ ⩽𝓒 𝕀CS and ℍ ∘ 𝕂 ⩾𝓘 𝕀IS. That is to say, for each (X, 𝓒) ∈ |CS| and (X, 𝓘) ∈ |IS|, 𝓒𝓘𝓒𝓒 𝓒 and 𝓘 ⩽𝓘 𝓘𝓒𝓘.

On one hand, take any x, yX. Then

J(x,y)x,yACJA=coCJ({x,y})=JCJ(x,y).

On the other hand, take any A ∈ 2X. Then

ACx,yA,coC({x,y})Ax,yA,JC(x,y)AACJC.

This means that 𝓒 ⊆ 𝓒𝓘𝓒, that is, 𝓒𝓘𝓒𝓒 𝓒, as desired.□

By Propositions 5.2 and 5.4, we know 𝕂* ≜ 𝕂 : ISCS(2) and ℍ* ≜ ℍ|CS(2) : CS(2)IS are still functors. Moreover, we have the following result.

Theorem 5.7

(𝕂*, ℍ*) is an adjunction between IS and CS(2). Moreover, 𝕂* is a left inverse of*.

Proof

By Theorem 5.6, it suffices to show that 𝓒𝓘𝓒 = 𝓒 for each arity 2 convex space (X, 𝓒). Take any A ∈ 2X. Then

ACx,yA,coC({x,y})A((X,C)is arity 2)x,yA,JC(x,y)AACJC.

This means 𝓒𝓘𝓒 = 𝓒.□

Corollary 5.8

The category CS(2) can be embedded in the category IS as a reflective subcategory.

6 Conclusions

In this paper we provided a categorical approach to abstract convex theory. On one hand, we introduced the axiomatic conditions of hull operators and showed that the resulting category is isomorphic to the category of convex spaces. On the other hand, we investigated the relations between convex spaces and interval spaces. We showed that there is an adjunction between the category of interval spaces and the category of convex spaces. Furthermore, the category of arity 2 convex spaces can be embedded in the category of interval spaces as a reflective subcategory. As is shown in this paper, category theory is an effective tool to deal with convex structures and interval operators. This also implies that category theory will be significant in the research on the theory of convex structures. In the future, we will consider applying category theory to fuzzy convex structures and establishing the relations between fuzzy convex structures and some other related structures.

Acknowledgement

This work is supported by the the Filling Project of Education Department of Heilongjiang Province (NO. 1352MSYYB008), the Scientific Research Project of Mudanjiang Normal University (NO. GP2017001), the Project of Shandong Province Higher Educational Science and Technology Program (NO. J18KA245).

References

About the article

Received: 2018-08-30

Accepted: 2019-02-20

Published Online: 2019-05-16


Citation Information: Open Mathematics, Volume 17, Issue 1, Pages 374–384, ISSN (Online) 2391-5455, DOI: https://doi.org/10.1515/math-2019-0029.

Export Citation

© 2019 Wang et al., published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 Public License. BY 4.0

Comments (0)

Please log in or register to comment.
Log in