• Aucun résultat trouvé

The Lattice of all Betweenness Relations: Structure and Properties

N/A
N/A
Protected

Academic year: 2022

Partager "The Lattice of all Betweenness Relations: Structure and Properties"

Copied!
9
0
0

Texte intégral

(1)

Struture and properties

LaurentBeaudou,MamadouMoustaphaKanté,andLhouariNourine

ClermontUniversité,UniversitéBlaisePasal,LIMOS,CNRS,Frane

laurent.beaudouuniv-bplermont. fr, {mamadou.kante,nourine}isim a.fr

Abstrat. Weonsiderimpliationbaseswithpremisesofsizeexatly

2

, whih are alsoknownas betweenness relations. Ourmotivations is thatseveralproblemsingraphtheoryanbemodelledusingbetweenness

relations,e.g.hullnumber,maximalliques.Inthispaperweharaterize

thelattieofallbetweennessrelationsbygivingitsposetofirreduible

elements.Moreover,weshowthatthislattieisameet-sublattieofthe

lattieofalllosuresystems.

1 Introdution

A onvexity spae on a ground set

X

is a subset of

2 X

that is losed under

intersetion. Convexity spaes were studied in [13℄ and are sometimes alled

Closuresystems.Themembersofaonvexityspaearealledonvexsets.Sine

the paper [13℄, onvexity spaes are studied by several authors who desribe

severaloftheirproperties(seethejointpaper[8℄ofEdelmanandJamisonfora

listofpubliationsduringtheeighties),inpartiularthesetofonvexityspaes

formsalattie.

Inthispaperwe dealwith betweenness relationswhihare speial asesof

onvexityspaes.Thenotionof betweennessrelationhasappearedintheearly

twentiethentury whenmathematiians fousedon fundamental geometry[4℄.

A betweenness relation

B

on a nite set

X

is a set of triples

(x, y, z ) ∈ X 3

.

The mostintuitive betweennessrelationsare thoseomingfrommetri spaes

(apoint

y

is between

x

and

z

if theysatisfythetriangularequality).Aonvex

set ofabetweennessrelation

B

isasubset

Y

of

X

suhthatforall

(x, y, z ) ∈ B

,

if

{x, z} ⊆ Y

, then

y ∈ Y

. It is well-known that the set of onvex sets of

a betweenness relation is a onvexity spae. Betweenness relations have been

thoroughlystudiedbyMengerandhisstudents[16℄.Otherbetweennessrelations

haveariseninresearheldsasprobabilitytheorywiththeworkofReihenbah

[18℄.Betweennessrelationshavebeenalsostudiedingraphsinordertogeneralize

geometrial theorems [1,2,9℄ (see the survey [17℄ for a non exhaustive list of

betweennessrelationsongraphs).

We are interested in desribing the set of all betweenness relations on a

groundset

X

.Weprovethatthesetof allonvexityspaeson

X

derivedfrom

betweennessrelationson

X

formsalattieandisinfatameet-sublattieofthe lattie ofallonvexity spaeson

X

(wegiveanexample showingthatit isnot

c 2012 by the paper authors. CLA 2012, pp. 317–325. Copying permitted only for private and academic purposes. Volume published and copyrighted by its editors.

Local Proceedings in ISBN 978–84–695–5252–0,

Universidad de M´ alaga (Dept. Matem´ atica Aplicada), Spain.

(2)

asub-lattie).Wealsodesribethe setofmeet andjoin-irreduibleelementsof

the lattie. We onlude by showing thatthe set of onvexity spaesobtained

fromlique betweenness relationsongraphs isa sublattie of the lattie ofall

onvexityspaesofbetweennessrelations.

Thispaperis motivatedby understandinglinksbetweenseveralparameters

that areonsidered in dierent areassuhasFCA,database, logi andgraph

theory.

Summary. NotationsanddenitionsaregiveninSetion2.Thedesriptionof

thelattieofonvexityspaesofbetweennessrelationsisgiveninSetion3.The

onvexityspaesofliquebetweennessrelationsisdesribedinSetion4.Some

questionsarisingfromalgorithmiaspetsaregiveninSetion5.

2 Preliminaries

Let

X

benite set.A partially ordered set on

X

(orposet) isareexive,anti-

symmetriandtransitivebinaryrelationdenotedby

P := (X, ≤)

.For

x, y ∈ X

,

we saythat

y

overs

x

,denotedby

x ≺ y

,ifforany

z ∈ X

with

x ≤ z ≤ y

we

have

x = z

or

y = z

.A lattie

L := (X, ≤)

is apartially ordered setwith the

followingproperties:

1. forall

x, y ∈ X

there existsaunique

z

,denoted by

x ∨ y

,suhthatforall

t ∈ X

,

t ≥ x

and

t ≥ y

implies

z ≤ t

.(Upperbound property.)

2. forall

x, y ∈ X

there existsaunique

z

,denoted by

x ∧ y

,suhthatforall

t ∈ X

,

t ≤ x

and

t ≤ y

implies

z ≥ t

.(Lowerboundproperty.)

Let

L = (X, ≤)

be a lattie. An element

x ∈ X

is alled join-irreduible (resp.meet-irreduible) if

x = y ∨ z

(resp.

x = y ∧ z

) implies

x = y

or

x = z

.

A join-irreduible(resp.meet-irreduible) element overs(resp. is overedby)

exatlyoneelement.Wedenoteby

J L

and

M L

thesetofalljoin-irreduibleand meet-irreduibleelementsof

L

respetively.

Theposetofirreduibleelementsofalattie

L = (X, ≤)

isarepresentation of

L

byabipartiteposet

Bip(L) = (J L , M L , ≤)

.Theoneptlattieof

Bip(L)

is isomorphi to

L

(for moredetails see the books of DaveyandPriestley [3℄,

andGanterandWille [10℄).

Animpliationon

X

isanordered pair

(A, B)

of subsetsof

X

,denotedby

A → B

. The set

A

is alled the premiseand the set

B

the onlusion of the

impliation

A → B

.Let

Σ

be aset of impliationson

X

.A subset

Y ⊆ X

is

Σ

-losedifforeahimpliation

A → B

in

Σ

,

A ⊆ Y

implies

B ⊆ Y

.Thelosure

ofaset

S

by

Σ

,denotedby

S Σ

,isthesmallest

Σ

-losedsetontaining

S

.

Let

S

beasubsetof

X

.Algorithm1omputesthelosureof

S

byabetween-

ness relation

Σ

.It is knownas forward haining proedure or hase proedure

[11℄.

Thesetof

Σ

-losedsubsetsof

X

,denotedby

F Σ

,isalosure system on

X

(3)

Algorithm1:SetClosure(

S, Σ

)

Data:Aset

S ⊆ X

and

Σ

abetweenness Result:Thelosure

S Σ

begin

Let

S Σ := S

;

while

∃xy → z ∈ Σ s.t. {x, y} ⊆ S Σ

and

z / ∈ S Σ

do

S Σ = S Σ ∪ {z}

;

end

Conversely,givenalosuresystem

F

on

X

,afamily

Σ

ofimpliationson

X

isalledanimpliationalbasisfor

F

if

F = F Σ

.Asubset

K ∈ X

isalledakey

if

K Σ = X

and

K

isminimalunderinlusionwiththisproperty.Thenamekey

omesfromdatabasetheory[15℄.

Denition 1. Animpliationset

Σ

on

X

isalledabetweennessrelationiffor all

A → B ∈ Σ

,

|A| = 2

.

Twobetweennessrelations

Σ 1

and

Σ 2

aresaidtobeequivalent,denotedby

Σ 1 ≡ Σ 2

,if

F Σ 1 = F Σ 2

.We denethe losure of abetweennessrelation

Σ

by

Σ c = {ab → c | a, b, c ∈ X

and

Σ ≡ Σ ∪ {ab → c}}

. Note that

Σ c

is the

uniquemaximalbetweennessrelationequivalentto

Σ

.Ineahequivalenelass

wedistinguishtwotypesofbetweennessrelations:

Canonial Aanonialbetweenness isthemaximuminitsequivalenelass.

Optimal Abetweenness

Σ

isoptimal ifforanybetweennessrelation

Σ

equiv-

alentto

Σ

,wehave

|Σ| ≤ |Σ |

.

A graph

G

is a pair

(V (G), E(G))

, where

V (G)

is the set of verties and

E (G)

is theset ofedges.Weonsidersimplegraphs(forfurtherdenitionssee

the book[6℄).Examplesofbetweennessrelationsarisingfromgraphtheoryare

Σ G = {xy → z | z

liesinashortestpathfrom

x

toy

}

and

Σ G = {xy → V (G) | xy ∈ E(G)}

.Severalothernotionsofonvexityspaesaredenedongraphs(see

[17℄). Figure1givesanexampleofagraph anditsonvexsetsforthe shortest

pathbetweenness.

3 The Lattie of all Betweenness Relations

Let

X

beanitesetand

Σ

abetweennessrelationon

X

.Giventwosets

A

and

C

in

2 X

suhthat

A ⊆ C

,wedenethe setinterval

[A, C]

asthe familyofall

sets

B

in

2 X

suhthat

A ⊆ B ⊆ C

.

Demetrovisetal.[5℄gaveaharaterizationofonvexsetsofanimpliation

(4)

a b c

d

(a)Agraph

G

{}

{a,c}

{a}

{b,d}

{b,c}

{a,d}

{a,b,c,d}

{c} {d} {b}

(b) Convex sets of

G

for

shortestpathbetweenness

Fig.1.Agraphanditsonvexsetsfortheshortestpathbetweennessrelation

Proposition1. [5 ℄Let

Σ

bea betweennessrelation onaset

X

.Then,

F Σ = 2 X \ [

ab→c∈Σ

[{a, b}, X \ {c}].

WedenotebyF

X := {F Σ | Σ

is abetweennessrelationon

X}

thefamilyof

allbetweennessrelationson

X

.

Theorem1. F

X

is a losure system and therefore a lattie when strutured

underinlusion.

Proof. Wehavetoprovethatthisstrutureislosedundertheintersetionand

itontainsauniquemaximalelement.

Let

F 1 , F 2 ∈

F

X

,thenthereexist

Σ 1

and

Σ 2

induingthesefamiliesofonvex

setson

X

.Let

F = F 1 ∩ F 2

and

Σ

bethebetweennessdenedby

ab → c ∈ Σ

if

ab → c ∈ Σ 1

or

ab → c ∈ Σ 2

.Thenwelaimthat

F = F Σ

.

Let

C

beasetin

F

.Itisonvexfor

Σ 1

andfor

Σ 2

.Therefore,forany

a, b

in

C

,every

c

suhthat

ab → c ∈ Σ 1

or

ab → c ∈ Σ 2

isin

C

.Fromthis,wederive

thatforany

a, b

in

C

,every

c

suhthat

ab → c ∈ Σ

isin

C

,sothat

C

isonvex

for

Σ

.

Reiproally,let

C

beaonvexsetfor

Σ

.Wewillshowthatitisonvexfor

Σ 1

and

Σ 2

.Let

a, b, c

beelementsof

X

suhthat

a

and

b

arein

C

and

ab → c ∈ Σ 1

.

Then

ab → c ∈ Σ

andsine

C

isonvexfor

Σ

,

b

is in

C

.Therefore,

C

is in

F 1

.

Similarlyitisin

F 2

sothatitis in

F

.

Sine

Σ = Σ 1 ∪ Σ 2

,weonludethat

Σ

isabetweennessrelationandthere- forethesetF

X

islosedunderintersetion.

Thefamily

2 X

isinF

X

.Itariseswhen

Σ

istheemptybetweennessrelation.

Proposition2. Given two families

F 1

and

F 2

in F

X

and their anonial be-

tweennessrelations

Σ 1

and

Σ 2

.Then

(5)

F 1 ∧ F 2 = F Σ 1 ∪Σ 2

.

F 1 ∨ F 2 = F Σ 1 ∩Σ 2

.

Proof. See theproofofTheorem1fortherstpoint.Noww,notethat

Σ 1

and

Σ 2

areanonialbetweennessrelations.Supposethat

F 1 ∨ F 2

isnot

F Σ 1 ∩Σ 2

.By

Proposition1,wehave

F 1 ∨ F 2 ⊆ F Σ 1 ∩Σ 2

.

Nowall

Σ ∨

the anonialbetweennessrelationrelatedto

F 1 ∨ F 2

.If

ab → c ∈ Σ ∨

for some

a, b

and

c

in

X

, then we know that

ab → c ∈ Σ i

beause

F i ⊆ F 1 ∨ F 2

for

i = 1, 2

(in the anonial form,we have every

ab → c

suh

that the orresponding interval has no intersetion with

F

). Therefore, every

ab → c ∈ Σ ∨

istruefor

Σ

,and

F Σ 1 ∩Σ 2 = F 1 ∨ F 2

.

Corollary1. F

X

isameet-sublattieofthelattieof alllosuresystemson

X

.

Proof. Let

F 1

and

F 2

betwofamiliesinF

X

.Sine

F 1 ∩ F 2

isthelosuresystem

of abetweennessrelationthen themeetis preservedandthusF

X

is aameet-

sublattie ofthe lattieofalllosuresystemson

X

.

Remark 1. NotiethatF

X

isnotasublattieofthelattieofalllosuresystems

on

X

.It sues toonsider the example where

X = {1, 2, 3, 4}

.Takethe o-

atoms

F 1 = 2 X \ [{1, 2}, {1, 2, 3}]

denedbythe betweennessrelationrestrited to

12 → 4

and

F 2 = 2 X \ [{2, 3}, {1, 2, 3}]

denedbythe betweennessrelation restrited to

23 → 4

.Then

F 1 ∪ F 2 = 2 X \ {{1, 2, 3}}

andis alosuresystem,

while

F 1 ∨ F 2

isthetopelement,

2 X

.

Inthefollowing,wegiveaharaterizationoftheposetofirreduibleelements

ofF

X

.

Proposition3. The poset of irreduible elements of F

X

is the bipartite poset

Bip(

F

X ) = (J

F

X , M

F

X , ⊆)

where

J

F

X := {F ⊥ ∪ {S} | S ∈ 2 X \ F ⊥ }

where

F ⊥ = {∅, X} ∪ {{x} | x ∈ X}

M

F

X := {2 X \ [ab, X \ {c}] | a, b, c ∈ X}.

Proof. Weprovethispropositionpoint bypoint.

Consider the maximal betweenness relation

Σ = {ab → c | a, b, c ∈ X}

.

Then

F ⊥ = F Σ = 2 X \ S

ab→c∈Σ [{a, b}, X \ {c}]

(see Proposition 1). Thus

F ⊥ = {∅, X} ∪ {{x} | x ∈ X}

.

For meet-irreduible elements, we will rst onsider o-atoms. Let

Σ

be a

betweennessrelation suh that

F Σ

is a o-atom. Sine

Σ

is non-empty, then

there exists a set whih is not onvex. Thus

Σ

must ontain an impliation

ab → c

,with

a, b, c ∈ X

,whihorrespondstothe maximallosuresystem in

F

X

anddierent from

2 X

.Namely,we removefrom

2 X

the onvexsetsof the

interval

[{a, b}, X \ {c}]

.

Nowsupposethereexistsanothermeet-irreduibleelement

F

whihisnota

o-atom. Call

Σ 1

a betweennessrelationsuh that

F

is

F Σ 1

. Also,all

F

the

(6)

only suessorof

F Σ 1

and

Σ 2

abetweennesssuhthat

F

is

F Σ 2

.By Proposi-

tion 1,weknowthatfrom

F

to

F

,we removeatleastanintervalofthe form

[{a, b}, X \ {c}]

.Thus, theo-atom assoiatedtothe impliation

ab → c

isnot

above

F

butitis above

F

,sothat

F

has atleasttwosuessors.Weonlude

thatallmeet-irreduibleelementsareo-atomsofF

X

.

For join-irreduible elements, we will rst haraterizeatoms of F

X

. Pik

any subset

S

of

X

whihis not empty,a singletonor the whole of

X

. Dene

thebetweennessrelation

Σ S

suhthat

ab → c ∈ Σ S

forevery

a, b, c

in

X

exept

those where

a

and

b

are in

S

and

c

is not in

S

.Then

F Σ S

is

F ⊥ ∪ {S}

.Now

supposethereexistsajoin-irreduible

F ∈

F

X

thatis notanatom.

F

ontains

at least one set

S

whih is not in the unique losure system

F

F

X

that

it overs. But there is an atom whihontains exatly

F ” = F ⊥ ∪ {S}

, with

F ” ⊆ F

and

F ” 6⊆ F

.Thus

F

oversatleasttwoelements,andthus

F

isnota

join-irreduibleelement.

Corollary2. F

X

ontains

n 2

(n − 2)2 n−3

meet-irreduible and

2 n − (n + 2)

join-irreduibleelements.

Proof. Everyo-atomisoftheform

2 X \[{a, b}, X \{c}]

andisaboveeveryatom

formedby

F ⊥

andanysetnotinthe forbiddeninterval.Thismakes

2 n − (n + 2) − 2 n−3

atomsbelowit.

Anatomoftheform

F ⊥ ∪ S

where

S

isaseton

X

ofsize2to

n − 1

,isbelow

every o-atom thatdoes not forbid

S

. This number equals

n 2

(n − 2)2 n−3 − [ |S| 2

(n − |S|)]

.

Figure2showstheirreduibleposetfor

X = {1, 2, 3, 4}

whereeveryatomis

represented bythe set

S

addedto

F ⊥

andeveryo-atomis representedby the removedimpliation

xy → z

.

12 → 3 12 → 4 13 → 2 13 → 4 14 → 2 14 → 3 23 → 1 23 → 4 24 → 1 24 → 3 34 → 1 34 → 2

{1, 2} {1, 3} {1, 4} {2, 3} {2, 4} {3, 4} {1, 2, 3} {1, 2, 4} {1, 3, 4} {2, 3, 4}

Fig.2.Irreduibleposetfor

n = 4

(7)

4 Clique Betweenness Relations

In this setion we deal with a speial betweenness relation dened througha

graph in a spei way. Many other betweenness relations on graphs an be

denedinthesameway,e.g.independentsets,setovers.

Givenagraph

G

,wedenethebetweennessrelation

Σ G := {ab → c | c ∈ X, ab / ∈ E(G)}.

The onvexsetsof

Σ G

areexatlythe liquesof

G

.Notie thatforanygraph

G

,thereisaorrespondingbetweennessrelation

Σ G

.Inthe following,wehar-

aterizethe lattieofall

Σ G

where

G

isa graphwithvertex-set

X

.Wedenote

byF

K X = {F Σ G | G

isagraphon

X}

.

Proposition4.

(

F

K X , ⊆)

isalattie.

Proof. Thebottom(resp.top)elementofF

X

orrespondsto

F Σ G

where

G

isa

stable(resp.lique)on

X

.

Moreover,F

K X

islosedunderintersetion(theliquesof theintersetionof twographsareexatlythe onesinthe intersetionofboth familiesof liques).

Therefore,

(

F

K X , ⊆)

isalattie.

Sine

Σ G

isabetweennessrelation,thenF

K X ⊂

F

X

forany graph

G

dened

on

X

.

Proposition5. Thelattie

(

F

K X , ⊆)

isa sublattieof

(

F

X , ⊆)

.

Proof. It is easy tosee thatitis ameet-sublattie,the meetof twofamiliesis

the intersetionofbothfamiliesinbothstrutures.

Inordertoprovethatitisajoin-sublattieof

(

F

X , ⊆)

,onsidertwographs

G 1

and

G 2

andtheirliquefamilies

F 1

and

F 2

.Call

G

thegraphunionof

G 1

and

G 2

and

F

thefamilyofitsliques.Therelatedbetweennessrelationisobtained

bytakingthe intersetionof betweennessrelationsrelatedto

G 1

and

G 2

(non-

edgesin

G

areexatlynon-edgesin

G 1

andin

G 2

).Thereforeitisthejoinof

F 1

and

F 2

in

(

F

X , ⊆)

.

Corollary3. Thelattie

(

F

K X , ⊂)

isabooleanlattie with

n 2

atoms.

Proof. Foreahgraph

G

,itsorrespondinganonialbetweennessrelationisthe set

Σ G c := {ab → X | ab / ∈ E(G)}

.Notethatanysuper-setof

Σ G c

orrespondsto

abetweennessrelationofapartialgraphof

G

,bydeletingedges

ab

from

G

whih

orrespondstoadding

ab → X

in

Σ G c

.Sineanyatomof

(

F

K X , ⊂)

orrespondsto

abetweennessrelationwhihontainsexatlyanimpliation

ab → X

,we have

exatly

n 2

atoms.

(8)

5 AlgorithmiAspets of Betweenness Relations

In this setion, we reall some optimization problems related to betweenness

relations.

MinimumKey(MK)

Input:

Σ

abetweennessrelationon

X

and

k

aninteger.

Question:Isthereaset

K ⊆ X

suhthat

|K| ≤ k

and

K Σ = X

?

Theseproblems have been studiedin the severaldomains andspeially in

database theory,andthey have been provedNP-omplete [5,14,15℄for general

impliationbases.TheproblemMKhasbeenprovedNP-ompleteforpartiular

asesofbetweennessrelations(seeforinstane[7℄fortheshortestpathbetween-

nessrelationongraphs).Itisknownasthehullnumberofabetweennessrelation.

Therefore,wehavethefollowing.

Proposition6. MK isNP-omplete.

Reently,Kanté andNourine [12℄ have shown that

M K

is polynomial for

shortest path betweennessrelationsof hordal anddistane hereditary graphs

byusingdatabasetehniques.CanweusethelattiestrutureofF

X

togetnew

polynomialtimealgorithmsfortheMKprobleminnewgraphlasses?

Nowweonsidertheproblemwhihomputesanoptimaloverofabetween-

ness relation.Thisproblemis tond anoptimalbetweennessrelationwhihis

equivalenttoagivenbetweennessrelation.Severalworkshavebeendoneinthe

generalaseknownasHorn minimization[11℄.

Optimal Cover(OC)

Input:

Σ

abetweennessrelationon

X

and

k

aninteger.

Question:Isthereabetweennessrelation

Σ

equivalentto

Σ

suhthat

| ≤ k

?

Thesizeofanoptimaloverisknownasthehydranumber [19℄.Theompu-

tationalomplexityof thehydranumberis openforbetweennessrelations,but

is NP-omplete forthegeneralase [11,15℄. Wehope thatthelattie struture

ofF

X

ouldhelptoaddressthisquestion.

6 Conlusion

In thispaper, we haraterizethe ontextof the lattie of all betweennessre-

lations on a nite set, whih is a meet-sublattie of the lattie of all losure

systemsonthesameset.Weareonvinedthatthestrutureoflattieanhelp

tounderstand someproblemsof graph theorysuhashullnumber andhydra

number.Inthe future wewillinvestigatethe linkof theseparameters andthe

(9)

Referenes

1. XiaominChenandVa²ekChvátal.ProblemsrelatedtoadeBruijn-Erd®stheorem.

DisreteAppliedMathematis,156(11):21012108,2008.

2. Va²ekChvátal. Antimatroids,betweenness,onvexity. InCook,W.J.,Lováz,L.,

Vygen, J. (eds.) Researh Trends in Combinatorial Optimization, pages 5764,

2019.

3. B. A. Daveyand A. Priestley. Introdution to Latties and Order. Cambridge

UniversityPress,2002.

4. G.deBeauregardRobinson. Thefoundationsofgeometry. Mathematialexposi-

tions.UniversityofTorontoPress,1952.

5. JánosDemetrovis,LeonidLibkin,andIlyaB.Muhnik.Funtionaldependenies

in relational databases: A lattie point of view. Disrete Applied Mathematis,

40(2):155185,1992.

6. ReinhardtDiestel. GraphTheory. Springer-Verlag,

3 rd

edition,2005.

7. Mitre Costa Dourado, John G. Gimbel, Jan Kratohvíl, Fábio Protti, and

JaymeLuizSzwarter. Ontheomputationofthehullnumberofagraph. Dis-

reteMathematis,309(18):56685674,2009.

8. P.H.EdelmanandR.E.Jamison. Thetheoryofonvexgeometries.InR.Rustin,

editor,Geometriae Dediata,vol19,Ni3,pages247270.springer,1985.

9. Martin Farber and Robert E. Jamison. Convexity ingraphs and hypergraphs.

SIAMJournalonAlgebraiandDisreteMethods,7:433444,1986.

10. B. Ganter and R. Wille. Formal onept analysis: Mathematial foundations.

Springer(BerlinandNewYork),1999.

11. PeterL.Hammer andAlexander Kogan. Optimal ompressionof propositional

hornknowledgebases: Complexityandapproximation. Artif.Intell.,64(1):131

145,1993.

12. M.M.KantéandL.Nourine.Polynomialtimealgorithmsforomputingaminimum

hullsetindistane-hereditaryandhordalgraphs. Submitted,2012.

13. D.C. KayandE.W. Womble. Axiomationvexitytheoryand relationships be-

tweenthe Carathéodory, Helly,and Radonnumbers. PaiJournal of Mathe-

matis,38:471485,1971.

14. ClaudioL.LuhesiandSylviaL.Osborn. Candidatekeysfor relations. Journal

ofComputerandSystemSienes,17(2):270279,1978.

15. DavidMaier. Minimumoversinrelationaldatabasemodel. J.ACM,27(4):664

674,1980.

16. KarlMenger. Untersuhungenüberallgemeinemetrik. Mathematishe Annalen,

100:75163,1928. 10.1007/BF0144884 0.

17. Ignaio M. Pelayo. On onvexity in graphs. Tehni-

al report, Universitat Politènia de Catalunya, http://www-

ma3.up.es/users/pelayo/researh/Denitions.pdf,2004.

18. H.ReihenbahandM. Reihenbah. TheDiretionofTime. CaliforniaLibrary

ReprintSeries.UniversityofCaliforniaPress,1956.

19. DespinaStasi,RobertH.Sloan,andGyörgyTurán.Hydraformulasanddireted

hypergraphs:Apreliminaryreport. InISAIM,2012.

Références

Documents relatifs

Thus each mathematical symbol (G, L,. .) has a single definite meaning inside each exercise, but might have different meanings from an exercise to another.. It is not required to

Les expériences fondées sur des modèles de déficiences génétiques en autophagie ont permis de démontrer l’importance de ce processus dans l’homéostasie des lymphocytes B et T

Note: This examination consists of ten questions on one page. a) Considering two arbitrary vectors a and b in three-dimensional Cartesian space, how can the angle between

Aware of the grave consequences of substance abuse, the United Nations system, including the World Health Organization, has been deeply involved in many aspects of prevention,

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

Suppose R is a right noetherian, left P-injective, and left min-CS ring such that every nonzero complement left ideal is not small (or not singular).. Then R

The celebrated Tits Alternative states that if G is a finitely generated linear group over any field, then either G contains a non-abelian free subgroup, or it is virtually

 Majority (75%, n=12) selected an article where document type was in the Single Search display.. Majority of users