[FREE PDF sample] Mathematical foundation of fuzzy sets 1st edition hsien-chung wu ebooks

Page 1


Mathematical Foundation of Fuzzy Sets 1st Edition

Hsien-Chung Wu

Visit to download the full and correct content document: https://ebookmass.com/product/mathematical-foundation-of-fuzzy-sets-1st-edition-hsi en-chung-wu/

More products digital (pdf, epub, mobi) instant download maybe you interests ...

Strengthening of Concrete Structures Using Fiber Reinforced Polymers (FRP): Design, Construction and Practical Applications Hwai-Chung Wu

https://ebookmass.com/product/strengthening-of-concretestructures-using-fiber-reinforced-polymers-frp-designconstruction-and-practical-applications-hwai-chung-wu/

Mathematical

Methods of

Analytical Mechanics

1st Edition Henri Gouin

https://ebookmass.com/product/mathematical-methods-of-analyticalmechanics-1st-edition-henri-gouin/

The Chemical Transformations of C1 Compounds 1st Edition Xiao-Feng Wu

https://ebookmass.com/product/the-chemical-transformationsof-c1-compounds-1st-edition-xiao-feng-wu/

Hollow Fiber Membranes: Fabrication and Applications 1st Edition Tai-Shung Chung (Editor)

https://ebookmass.com/product/hollow-fiber-membranes-fabricationand-applications-1st-edition-tai-shung-chung-editor/

Operative Techniques : Hand and Wrist Surgery 3rd. Edition Chung

https://ebookmass.com/product/operative-techniques-hand-andwrist-surgery-3rd-edition-chung/

Transition Metal-Catalyzed Benzofuran Synthesis 1st Edition Xiao-Feng Wu

https://ebookmass.com/product/transition-metal-catalyzedbenzofuran-synthesis-1st-edition-xiao-feng-wu/

A Modern Introduction to Fuzzy Mathematics 1. Edition

Apostolos Syropoulos

https://ebookmass.com/product/a-modern-introduction-to-fuzzymathematics-1-edition-apostolos-syropoulos/

Multiphase Fluid Flow in Porous and Fractured Reservoirs 1st Edition Wu

https://ebookmass.com/product/multiphase-fluid-flow-in-porousand-fractured-reservoirs-1st-edition-wu/

Principles of foundation engineering Ninth Edition

Braja M. Das

https://ebookmass.com/product/principles-of-foundationengineering-ninth-edition-braja-m-das/

MathematicalFoundationsofFuzzySets

MathematicalFoundationsofFuzzySets

Hsien-ChungWu

DepartmentofMathematics

NationalKaohsiungNormalUniversity

Kaohsiung

Taiwan

Thiseditionfirstpublished2023 ©2023JohnWileyandSons,Ltd

Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,or transmitted,inanyformorbyanymeans,electronic,mechanical,photocopying,recordingorotherwise, exceptaspermittedbylaw.Adviceonhowtoobtainpermissiontoreusematerialfromthistitleisavailable athttp://www.wiley.com/go/permissions.

TherightofHsien-ChungWutobeidentifiedastheauthorofthisworkhasbeenassertedinaccordance withlaw.

RegisteredOffices

JohnWiley&Sons,Inc.,111RiverStreet,Hoboken,NJ07030,USA

JohnWiley&SonsLtd,TheAtrium,SouthernGate,Chichester,WestSussex,PO198SQ,UK

Fordetailsofourglobaleditorialoffices,customerservices,andmoreinformationaboutWileyproducts visitusatwww.wiley.com.

Wileyalsopublishesitsbooksinavarietyofelectronicformatsandbyprint-on-demand.Somecontentthat appearsinstandardprintversionsofthisbookmaynotbeavailableinotherformats.

Trademarks:WileyandtheWileylogoaretrademarksorregisteredtrademarksofJohnWiley&Sons,Inc. and/oritsaffiliatesintheUnitedStatesandothercountriesandmaynotbeusedwithoutwritten permission.Allothertrademarksarethepropertyoftheirrespectiveowners.JohnWiley&Sons,Inc. isnotassociatedwithanyproductorvendormentionedinthisbook.

LimitofLiability/DisclaimerofWarranty

Whilethepublisherandauthorshaveusedtheirbesteffortsinpreparingthiswork,theymakeno representationsorwarrantieswithrespecttotheaccuracyorcompletenessofthecontentsofthisworkand specificallydisclaimallwarranties,includingwithoutlimitationanyimpliedwarrantiesofmerchantability orfitnessforaparticularpurpose.Nowarrantymaybecreatedorextendedbysalesrepresentatives,written salesmaterialsorpromotionalstatementsforthiswork.Thefactthatanorganization,website,orproduct isreferredtointhisworkasacitationand/orpotentialsourceoffurtherinformationdoesnotmeanthat thepublisherandauthorsendorsetheinformationorservicestheorganization,website,orproductmay provideorrecommendationsitmaymake.Thisworkissoldwiththeunderstandingthatthepublisheris notengagedinrenderingprofessionalservices.Theadviceandstrategiescontainedhereinmaynotbe suitableforyoursituation.Youshouldconsultwithaspecialistwhereappropriate.Further,readersshould beawarethatwebsiteslistedinthisworkmayhavechangedordisappearedbetweenwhenthisworkwas writtenandwhenitisread.Neitherthepublishernorauthorsshallbeliableforanylossofprofitorany othercommercialdamages,includingbutnotlimitedtospecial,incidental,consequential,orother damages.

LibraryofCongressCataloging-in-PublicationDataAppliedfor: HardbackISBN:9781119981527

CoverDesign:Wiley

CoverImage:©oxygen/GettyImages

Setin9.5/12.5ptSTIXTwoTextbyStraive,Chennai,India

Contents

Preface ix

1MathematicalAnalysis 1

1.1InfimumandSupremum 1

1.2LimitInferiorandLimitSuperior 3

1.3Semi-Continuity 11

1.4Miscellaneous 19

2FuzzySets 23

2.1MembershipFunctions 23

2.2 �� -levelSets 24

2.3TypesofFuzzySets 34

3SetOperationsofFuzzySets 43

3.1ComplementofFuzzySets 43

3.2IntersectionofFuzzySets 44

3.3UnionofFuzzySets 51

3.4InductiveandDirectDefinitions 56

3.5 �� -LevelSetsofIntersectionandUnion 61

3.6MixedSetOperations 65

4GeneralizedExtensionPrinciple 69

4.1ExtensionPrincipleBasedontheEuclideanSpace 69

4.2ExtensionPrincipleBasedontheProductSpaces 75

4.3ExtensionPrincipleBasedontheTriangularNorms 84

4.4GeneralizedExtensionPrinciple 92

5GeneratingFuzzySets 109

5.1FamiliesofSets 110

5.2NestedFamilies 112

5.3GeneratingFuzzySetsfromNestedFamilies 119

5.4GeneratingFuzzySetsBasedontheExpressionintheDecomposition Theorem 123

5.4.1TheOrdinarySituation 123

5.4.2BasedonOneFunction 129

5.4.3BasedonTwoFunctions 140

5.5GeneratingFuzzyIntervals 150

5.6UniquenessofConstruction 160

6FuzzificationofCrispFunctions 173

6.1FuzzificationUsingtheExtensionPrinciple 173

6.2FuzzificationUsingtheExpressionintheDecompositionTheorem 176

6.2.1NestedFamilyUsing �� -LevelSets 177

6.2.2NestedFamilyUsingEndpoints 181

6.2.3Non-NestedFamilyUsingEndpoints 184

6.3TheRelationshipsbetweenEPandDT 187

6.3.1TheEquivalences 187

6.3.2TheFuzziness 191

6.4DifferentiationofFuzzyFunctions 196

6.4.1DefinedonOpenIntervals 196

6.4.2FuzzificationofDifferentiableFunctionsUsingtheExtensionPrinciple 197

6.4.3FuzzificationofDifferentiableFunctionsUsingtheExpressioninthe DecompositionTheorem 198

6.5IntegralsofFuzzyFunctions 201

6.5.1LebesgueIntegralsonaMeasurableSet 201

6.5.2FuzzyRiemannIntegralsUsingtheExpressionintheDecomposition Theorem 203

6.5.3FuzzyRiemannIntegralsUsingtheExtensionPrinciple 207

7ArithmeticsofFuzzySets 211

7.1ArithmeticsofFuzzySetsin ℝ 211

7.1.1ArithmeticsofFuzzyIntervals 214

7.1.2ArithmeticsUsingEPandDT 220

7.1.2.1AdditionofFuzzyIntervals 220

7.1.2.2DifferenceofFuzzyIntervals 222

7.1.2.3MultiplicationofFuzzyIntervals 224

7.2ArithmeticsofFuzzyVectors 227

7.2.1ArithmeticsUsingtheExtensionPrinciple 230

7.2.2ArithmeticsUsingtheExpressionintheDecompositionTheorem 230

7.3DifferenceofVectorsofFuzzyIntervals 235

7.3.1 �� -LevelSetsof ̃ A ⊖EP ̃ B 235

7.3.2 �� -LevelSetsof ̃ A ⊖⋄ DT ̃ B 237

7.3.3 �� -LevelSetsof ̃ A ⊖⋆DT ̃ B 239

7.3.4 �� -LevelSetsof A ⊖† DT B 241

7.3.5TheEquivalencesandFuzziness 243

7.4AdditionofVectorsofFuzzyIntervals 244

7.4.1 �� -LevelSetsof A ⊕EP B 244

7.4.2 �� -LevelSetsof A ⊕DT B 246

7.5ArithmeticOperationsUsingCompatibilityandAssociativity 249

7.5.1Compatibility 250

7.5.2Associativity 255

7.5.3ComputationalProcedure 264

7.6BinaryOperations 268

7.6.1FirstTypeofBinaryOperation 269

7.6.2SecondTypeofBinaryOperation 273

7.6.3ThirdTypeofBinaryOperation 274

7.6.4ExistenceandEquivalence 277

7.6.5EquivalentArithmeticOperationsonFuzzySetsin ℝ 282

7.6.6EquivalentAdditionsofFuzzySetsin ℝm 289

7.7HausdorffDifferences 294

7.7.1FairHausdorffDifference 294

7.7.2CompositeHausdorffDifference 299

7.7.3CompleteCompositeHausdorffDifference 304

7.8ApplicationsandConclusions 312

7.8.1GradualNumbers 312

7.8.2FuzzyLinearSystems 313

7.8.3SummaryandConclusion 315

8InnerProductofFuzzyVectors 317

8.1TheFirstTypeofInnerProduct 317

8.1.1UsingtheExtensionPrinciple 318

8.1.2UsingtheExpressionintheDecompositionTheorem 322

8.1.2.1TheInnerProduct A ⊛⋄ DT B 323

8.1.2.2TheInnerProduct A ⊛⋆DT B 325

8.1.2.3TheInnerProduct A ⊛† DT B 327

8.1.3TheEquivalencesandFuzziness 329

8.2TheSecondTypeofInnerProduct 330

8.2.1UsingtheExtensionPrinciple 333

8.2.2UsingtheExpressionintheDecompositionTheorem 335

8.2.3ComparisonofFuzziness 338

9GradualElementsandGradualSets 343

9.1GradualElementsandGradualSets 343

9.2FuzzificationUsingGradualNumbers 347

9.3ElementsandSubsetsofFuzzyIntervals 348

9.4SetOperationsUsingGradualElements 351

9.4.1ComplementSet 351

9.4.2IntersectionandUnion 353

9.4.3Associativity 359

9.4.4EquivalencewiththeConventionalSituation 363

9.5ArithmeticsUsingGradualNumbers 364

10DualityinFuzzySets 373

10.1LowerandUpperLevelSets 373

10.2DualFuzzySets 376

10.3DualExtensionPrinciple 378

10.4DualArithmeticsofFuzzySets 380

10.5RepresentationTheoremforDual-FuzzifiedFunction 385

Bibliography 389

MathematicalNotations 397

Index 401

Preface

Theconceptoffuzzyset,introducedbyL.A.Zadehin1965,triedtoextendclassicalset theory.Itiswellknownthataclassicalsetcorrespondstoanindicatorfunctionwhosevaluesareonlytakentobe0and1.Withtheaidofamembershipfunctionassociatedwitha fuzzyset,eachelementinasetisallowedtotakeanyvaluesbetween0and1,whichcan beregardedasthedegreeofmembership.Thiskindofimprecisiondrawsforthabunchof applications.

Thisbookisintendedtopresentthemathematicalfoundationsoffuzzysets,whichcan rigorouslybeusedasabasictooltostudyengineeringandeconomicsproblemsinafuzzy environment.Itmayalsobeusedasagraduateleveltextbook.Themainprerequisitesfor mostofthematerialinthisbookaremathematicalanalysisincludingsemi-continuities, supremum,convexity,andbasictopologicalconceptsofEuclideanspace, ℝn .Thisbook presentsthecurrentstateofaffairsinsetoperationsoffuzzysets,arithmeticoperations offuzzyintervalsandfuzzificationofcrispfunctionsthatarefrequentlyadoptedtomodel engineeringandeconomicsproblemswithfuzzyuncertainty.Especially,theconceptsof gradualsetsandgradualelementshavebeenpresentedinordertocopewiththedifficulty forconsideringelementsoffuzzysetssuchasconsideringelementsofcrispsets.

● Chapter1presentsthemathematicaltoolsthatareusedtostudytheessenceoffuzzy sets.Theconceptsofsupremumandsemi-continuityandtheirpropertiesarefrequently invokedtoestablishtheequivalencesamongthedifferentsettingsofsetoperationsand arithmeticoperationsoffuzzysets.

● Chapter2introducesthebasicconceptsandpropertiesoffuzzysetssuchasmembership functionsandlevelsets.Thefuzzyintervalsarecategorizedasdifferenttypesbasedon thedifferentassumptionsofmembershipfunctionsinordertobeusedforthedifferent purposesofapplications.

● Chapter3dealswiththeintersectionandunionoffuzzysetsincludingthecomplementof fuzzysets.Thegeneralsettingsbyconsideringaggregationfunctionshavebeenpresented tostudytheintersectionandunionoffuzzysetsthatcovertheconventionalonessuch asusingminimumandmaximumfunctions(t-normands-norm)forintersectionand union,respectively.

● Chapter4extendstheconventionalextensionprincipletotheso-calledgeneralized extensionprinciplebyusinggeneralaggregationfunctionsinsteadofusingminimumfunctionort-normtofuzzifycrispfunctions.Fuzzificationsofreal-valuedand vector-valuedfunctionsarefrequentlyadoptedinengineeringandeconomicsproblems thatinvolvefuzzydata,whichmeansthatthereal-valueddatacannotbeexactlycollected owingtothefluctuationofanuncertainsituation.

x Preface

● Chapter5presentsthemethodologyforgeneratingfuzzysetsfromanestedfamilyor non-nestedfamilyofsubsetsofEuclideanspace ℝn .Especially,generatingfuzzyintervalsfromanestedfamilyornon-nestedfamilyofboundedclosedintervalsisusefulfor fuzzifyingthereal-valueddataintofuzzydata.Basedonacollectionofreal-valueddata, wecangenerateafuzzysetthatcanessentiallyrepresentthiscollectionofreal-valued data.

● Chapter6dealswiththefuzzificationofcrispfunctions.Usingtheextensionprinciple presentedinChapter4canfuzzifycrispfunctions.Thischapterstudiesanothermethodologytofuzzifycrispfunctionsusingthemathematicalexpressioninthewell-known decompositiontheorem.Theirequivalencesarealsoestablishedundersomemild assumptions.

● Chapter7studiesthearithmeticoperationsoffuzzysets.Theconventionalarithmetic operationsoffuzzysetsarebasedontheextensionprinciplepresentedinChapter4.Many otherarithmeticoperationsusingthegeneralaggregationfunctionshavenalsobeenstudied.Theequivalencesamongthesedifferentsettingsofarithmeticoperationsarealso establishedinordertodemonstratetheconsistentusageinapplications.

● Chapter8givesacomprehensiveandaccessiblestudyregardinginnerproductoffuzzy vectorsthatcanbetreatedasanapplicationusingthemethodologiespresentedin Chapter7.Thepotentialapplicationsofinnerproductoffuzzyvectorsarefuzzylinear programmingproblemsandtheengineeringproblemsthatareformulatedusingthe formofinnerproductinvolvingfuzzydata.

● Chapter9introducestheconceptsofgradualsetsandgradualelementsthatcanbeused toproposetheconceptofelementsoffuzzysetssuchastheconceptofelementsofcrisp sets.Roughlyspeaking,afuzzysetcanbetreatedasacollectionofgradualelements.In otherwords,afuzzysetconsistsofgradualelements.Inthiscase,thesetoperationsand arithmeticoperationsoffuzzysetscanbedefinedastheoperationsofgradualelements, liketheoperationsofelementsofcrispsets.Theequivalenceswiththeconventionalset operationsandarithmeticoperationsoffuzzysetsarealsoestablishedundersomemild assumptions.

● Chapter10dealswiththeconceptofdualityoffuzzysetsbyconsideringthelower �� -level sets.Theconventional �� -levelsetsaretreatedasupper �� -levelsets.Thischapterconsidersthelower �� -levelsetsthatcanberegardedasthedualofupper �� -levelsets.The well-knownextensionprincipleanddecompositiontheoremarealsoestablishedbased onthelower �� -levelsets,andarecalledthedualextensionprincipleanddualdecompositiontheorem.Theso-calleddualarithmeticsoffuzzysetsarealsoproposedbasedon thelower �� -levelsets,andadualityrelationwiththeconventionalarithmeticsoffuzzy setsisalsoestablished.

Finally,Iwouldliketothankthepublisherfortheircooperationintherealizationof thisbook.

DepartmentofMathematics

NationalKaohsiungNormalUniversity Kaohsiung,Taiwan

e-mail1:hcwu@mail.nknu.edu.tw

e-mail2:hsien.chung.wu@gmail.com

Website:https://sites.google.com/view/hsien-chung-wu April,2022

MathematicalAnalysis

Wepresentsomematerialsfrommathematicalanalysis,whichwillbeusedthroughoutthis book.Moredetailedargumentscanbefoundinanymathematicalanalysismonograph.

1.1InfimumandSupremum

Let S beasubsetof ℝ.Theupperandlowerboundsof S aredefinedbelow.

● Wesaythat u isan upperbound of S whenthereexistsarealnumber u satisfying x ≤ u forevery x ∈ S.Inthiscase,wealsosaythat S isboundedaboveby u

● Wesaythat l isa lowerbound of S whenthereexistsarealnumber l satisfying x ≥ l for every x ∈ S.Inthiscase,wealsosaythat S isboundedbelowby l

Theset S issaidtobeunboundedabovewhentheset S hasnoupperbound.Theset S is saidtobeunboundedbelowwhentheset S hasnolowerbound.Themaximalandminimal elementsof S aredefinedbelow.

● Wesaythat u∗ isa maximalelement of S whenthereexistsarealnumber u∗ ∈ S satisfying x ≤ u∗ forevery x ∈ S.Inthiscase,wewrite u∗ = max S.

● Wesaythat l∗ isa minimalelement of S whenthereexistsarealnumber l∗ ∈ S satisfying x ≥ l∗ forevery x ∈ S.Inthiscase,wewrite l∗ = min S.

Example1.1.1 Weprovidesomeconcreteexamples.

(i)Theset ℝ+ =(0, +∞) isunboundedabove.Ithasnoupperboundsandnomaximal element.Itisboundedbelowby0,butithasnominimalelement.

(ii)Theclosedinterval S =[0,1] isboundedaboveby1andisboundedbelowby0.Wealso havemax S = 1andmin S = 0.

(iii)Thehalf-openinterval S =[0,1) isboundedaboveby1,butithasnomaximalelement. However,wehavemin S = 0.

Althoughtheset S =[0,1) isboundedaboveby1,ithasnomaximalelement.Thismotivatesustointroducetheconceptsofsupremumandinfinum.

Definition1.1.2 Let S beasubsetof ℝ.

(i)Supposethat S isboundedabove.Arealnumber u ∈ ℝ iscalleda leastupperbound or supremum of S whenthefollowingconditionsaresatisfied.

● u isanupperboundof S.

● If u isanyupperboundof S,then u ≥ u. Inthiscase,wewrite u = sup S.Wesaythatthesupremumsup S is attained when u ∈ S.

(ii)Supposethat S isboundedbelow.Arealnumber l ∈ ℝ iscalleda greatestlower bound or infimum of S whenthefollowingconditionsaresatisfied.

● l isalowerboundof S.

● If l isanylowerboundof S,then l ≤ l. Inthiscase,wewrite l = inf S.Wesaythattheinfimuminf S is attained when l ∈ S.

Itiscleartoseethatifthesupremumsup S isattained,thenmax S = sup S.Similarly, iftheinfimuminf S isattained,thenmin S = inf S

Example1.1.3 Let S =[0,1].Then,wehave max S = sup S = 1andinf S = min S = 0.

If S =[0,1),thenmax S doesnotexists.However,wehavesup S = 1.

Proposition1.1.4 LetSbeasubsetof ℝ with u = sup S.Then,givenanys < u,thereexists t ∈ Ssatisfyings < t ≤ u.

Proof. Wearegoingtoproveitbycontradiction.Supposethatwehave t ≤ s forall t ∈ S Then s isanupperboundof S.Accordingtothedefinitionofsupremum,wealsohave s ≥ u Thiscontradictionimpliesthat s < t forsome t ∈ S,andtheproofiscomplete. ◾

Proposition1.1.5 GivenanytwononemptysubsetsAandBof ℝ,wedefineC = A + Bby

C = {x + y ∶ x ∈ Aandy ∈ B} .

Supposethatthesupremum sup Aand sup Bareattained.Then,thesupremum sup Cis attained,andwehave sup C = sup A + sup B.

Proof. Wefirsthave sup A = max A andsup B = max B

Wewrite a = sup A and b = sup B.Givenany z ∈ C,thereexist x ∈ A and y ∈ B satisfying z = x + y.Since x ≤ a and y ≤ b,wehave z = x + y ≤ a + b,whichsaysthat a + b isanupper boundof C.Therefore,thedefinitionof c = sup C saysthat c ≤ a + b.Next,wewantto showthat a + b ≤ c.Givenany ��> 0,Proposition1.1.4saysthatthereexist x ∈ A and y ∈ B satisfying a ��< x and b ��< y.Wealsoseethat x + y ≤ c.Addingtheseinequalities,we obtain

a + b 2��< x + y ≤ c, whichsaysthat a + b < c + 2�� .Since �� canbeanypositiverealnumber,wemusthave a + b ≤ c.Thiscompletestheproof. ◾

Proposition1.1.6 LetAandBbeanytwononemptysubsetsof ℝ satisfyinga ≤ bforany a ∈ Aandb ∈ B.Supposethatthesupremum sup Bisattained.Then,thesupremum sup A isattainedand sup A ≤ sup B.

Proof. Itisleftasanexercise. ◾

1.2LimitInferiorandLimitSuperior

Let {an }∞ n=1 beasequencein ℝ.The limitsuperior of {an }∞ n=1 isdefinedby limsup n→∞ an = inf n≥1 sup k≥n ak , andthe limitinferior of {an }∞ n=1 isdefinedby liminf n→∞ an =−limsup n→∞ (−an )

Moreover,wecanseethat

Let

Itiscleartoseethat {bn }∞ n=1 isadecreasingsequenceand {cn }∞ n=1 isanincreasingsequence. Inthiscase,wehave

whichalsosaysthat

and

Someusefulpropertiesaregivenbelow.

Proposition1.2.1 Let {an }∞ n=1 beasequenceofrealnumbers.Then,thefollowingstatementsholdtrue.

(i) Wehave

(ii) Wehave

ifandonlyif

an

(iii) Thesequencedivergesto +∞ ifandonlyif

(iv) Thesequencedivergesto −∞ ifandonlyif liminf n→∞ an = limsup n→∞ an =−∞.

(v) Let {bn }∞ n=1 beanothersequencesatisfyingan ≤ bn foralln.Then,wehave liminf n→∞ an ≤ liminf n→∞ bn and limsup n→

Proof. Toprovepart(i),from(1.1),weseethat cn ≤ bn forall n.Using(1.2)and(1.3),we obtain

liminf n→∞ an = lim n→∞ cn ≤ lim n→∞ bn = limsup n→∞ an .

Toprovepart(ii),supposethat

lim n→∞ an = a

Then,givenany ��> 0,thereexistsaninteger N satisfying

Inotherwords,wehave a �� 2 ≤ cn ≤ bn ≤ a

whichalsoimplies

Therefore,weobtain

∞ bn , whichimplies,byusing(1.2)and(1.3),

liminf n→∞ an = limsup n→∞ an = a

Fortheconverse,from(1.1)again,weseethat cn ≤ an ≤ bn forall n ≥ 1.Since

Usingthepinchingtheorem,weobtainthedesiredlimit.Theremainingproofsareleftas exercise,andtheproofiscomplete.

Proposition1.2.2 Let {an }∞ n=1 and {bn }∞ n=1 beanytwosequencesin ℝ.Then,wehave

Proof. For k ≥ n,wehave

+ bk ≤ sup

ak + sup

Therefore,weobtain

Wesimilarlyhave

Therefore,wealsoobtain

bk ] (using(1.5))

(sincethelimitsexist)

(using(1.2)and(1.3)).

Thiscompletestheproof. ◾

Proposition1.2.3 Let {An }∞ n=1 beasequenceofsubsetsof ℝm satisfyingAn+1 ⊆ An foralln and ⋂∞ n=1 An = A,andletfbeareal-valuedfunctiondefinedon ℝm .Then

and

and

Proof. Since inf a

Itsufficestoprovethecaseofthesupremum.Let

Since An+1 ⊆ An forall n,wehavethat {y∗ n }∞ n=1 isadecreasingsequenceofrealnumbers. Wealsohave y∗ n ≥ y∗ forall n,whichimplies liminf n→∞ y∗ n ≥ y∗ .

Givenany ��> 0,accordingtotheconceptofsupremum,thereexists an ∈ An satisfying y∗ n �� ≤ f (an )

Let bn = infk≥n f (ak ).Weconsiderthesubsequence {am }∞ m=1 definedby am = am+n 1 inthe senseof

{a1 , a2 , … , am , …} = {an , an+1 , … , am+n 1 , …} .

Then bn = infm≥n f (am ) and bn ≤ f (am ) forall m.Since An+1 ⊆ An forall n and ⋂∞ n=1 An = A, the“lastterm”ofthesequence {am }∞ m=1 mustbein A,aclaimthatwillbeprovedbelow. Since ak ∈ Ak ⊆ An forall k ≥ n,wehavethesubsequence {ak }∞ k=n ⊆ An ,whichalsoimplies A ≡ ∞ ⋂ n=1 ({ak }∞ k=n ) ⊆ ∞ ⋂ n=1 An = A,

where A canberegardedasthe“lastterm”and A ⊆ {am }∞ m=1 .Since y∗ isthesupremumof f on A,itfollowsthat f (a) ≤ y∗ foreach a ∈ A ⊆ A.Since bn ≤ f (am ) forall m,weseethat bn ≤ y∗ forall n.Therefore,weobtain

liminf n→∞ f (an )= sup n≥1 inf k≥n f (ak )= sup n≥1 bn ≤ y∗ , whichimplies,by(1.7),

liminf n→∞ y∗ n �� ≤ liminf n→∞ f (an ) ≤ y∗

Since �� isanypositivenumber,weobtain

liminf n→∞

Combining(1.6)and(1.8),weobtain sup n≥1 inf k≥n y∗ k = liminf n→∞ y∗ n = y∗ .

Since {y∗ n }∞ n=1 isadecreasingsequenceofrealnumbers,weconcludethat inf n

lim

liminf

andtheproofiscomplete.

(1.8)

Proposition1.2.4 Let {An }∞ n=1 beasequenceofsubsetsof ℝm satisfyingAn ⊆ An+1 foralln and ⋃∞ n=1 An = A,andletfbeareal-valuedfunctiondefinedon ℝm .Then

and

and

Proof. Itsufficestoprovethecaseofthesupremum.Let

y∗ n = sup a∈An f (a) and y∗ = sup a∈A f (a).

Since An ⊆ An+1 forall n,wehavethat {y∗ n }∞ n=1 isanincreasingsequenceofrealnumbers. Wealsohave y∗ n ≤ y∗ forall n,whichimplies

Givenany ��> 0,accordingtotheconceptofsupremum,thereexists a∗ ∈ A satisfying y∗ �� ≤ f (a∗ ).Since a∗ ∈ A = ∞ ⋃ n=1 An ,

wehavethat a∗ ∈ An∗ forsomeinteger n∗ .Weconstructasequence {an }∞ n=1 satisfying an ∈ An forall n < n∗ and an = a∗ forall n ≥ n∗ .Since An ⊆ An+1 forall n,itfollowsthat an ∈ An forall n ≥ n∗ .Therefore,thesequence {an }∞ n=1 satisfies an ∈ An forall n and a∗ ∈{ak }∞ k=n forall n, whichmeansthat a∗ isthe“lastterm”ofthesequence {an }∞ n=1 .Wealsohave y∗ n ≥ f (an ) (1.10)

Let bn = supk≥n f (ak ).Weconsiderthesubsequence {ap }∞ p=1 definedby ap = ap+n 1 inthe senseof

{a1 , a2 , … , ap , …} = {an , an+1 , … , ap+n 1 , …} .

Then bn = supp≥n f (ap ) and bn ≥ f (ap ) forall p.Since a∗ isthe“lastterm”ofthesequence {an }∞ n=1 ,itfollowsthat a∗ isalsothe“lastterm”ofthesequence {am }∞ m=1 .Therefore,we have bn ≥ f (a∗ ) ≥ y∗ �� forall n,whichimplies

limsup n→∞ f (an )= inf n≥1 sup k≥n f (ak )= inf n≥1 bn ≥ y∗ ��.

Since �� isanypositivenumber,itfollowsthat

limsup n→∞ f (an ) ≥ y∗ .

Using(1.10),weobtain limsup

Combining(1.9)and(1.11),weobtain inf n≥1 sup k

Since {y∗ n }∞ n=1 isanincreasingsequenceofrealnumbers,weconcludethat sup n

, andtheproofiscomplete.

Givenany x =(

and

) in ℝm .TheEuclideandistance between x and y isdefinedby ∥

Givenapoint x ∈ ℝm ,weconsidertheopen �� -ball B(x ; �� )= {y ∈ ℝm ∶∥ x y ∥<�� } . (1.12)

Theconceptofclosurebasedonopenballswillbefrequentlyusedthroughoutthisbook. ForthegeneralconceptrefertoKelleyandNamioka[55].Inthisbook,wearegoingto considertheclosureofasubsetof ℝm ,whichisgivenbelow.

Definition1.2.5 Let A beasubsetof ℝm .The closure of A isdenotedanddefinedby cl(A)= {x ∈ ℝm ∶ A ∩ B(x ; �� ) ≠ ∅ forany ��> 0}

Wesaythat A isaclosedsubsetof ℝm when A = cl(A).

Remark1.2.6 Givenany x ∈ cl(A),thereexistsasequence {xn }∞ n=1 in A satisfying ∥ xn x ∥→ 0as n → ∞.Inparticular,for m = 1,weseethat xn → x as n → ∞

Proposition1.2.7 LetAbeasubsetof ℝ,andletfbeacontinuousfunctiondefinedoncl(A). Then

sup a∈A f (a)= sup a∈cl(A) f (a) and inf a∈A f (a)= inf a∈cl(A) f (a).

Proof. Itsufficestoprovethecaseofthesupremum,since inf a∈A f (a)=−sup a∈A [−f (a)]. Itisobviousthat

sup a∈A f (a) ≤ sup a∈cl(A) f (a).

Givenany ��> 0,accordingtotheconceptofsupremum,thereexists a∗ ∈ cl(A) satisfying

sup a∈cl(A) f (a)− �� ≤ f (a∗ ) .

Wealsoseethatthereexistsasequence {an }∞ n=1 in A satisfying an → a∗ .Since f iscontinuousoncl(A),wealsohave f (an ) → f (a∗ ) as n → ∞.Therefore,weobtain sup a∈cl(A) f (a)− �� ≤ f (a∗ ) = lim n→∞ f (an ) ≤ lim n→∞ [sup a∈

Since �� canbeanypositivenumber,itfollowsthat

sup a∈cl(A) f (a) ≤ sup a∈A f (a)

Thiscompletestheproof. ◾

Let S beasubsetof ℝ.For a ∈ S andasequence {an }∞ n=1 in ℝ,wewrite an ↑ a tomean thatthesequence {an }∞ n=1 isincreasingandconvergesto a.Wealsowrite an ↓ a tomean thatthesequence {an }∞ n=1 isdecreasingandconvergesto a

Proposition1.2.8 LetAbeasubsetof ℝ.Thefollowingstatementsholdtrue.

(i) Letfbearight-continuousfunctiondefinedoncl(A).Givenanyfixedr ∈ ℝ,supposethat thereexistsasequence {an }∞ n=1 inAsatisfyingan ↓ rasn → ∞ andan > rforalln.Then, wehave

(ii) Letfbeacontinuousfunctiondefinedoncl(A).Givenanyfixedr ∈ ℝ,supposethatthere existsasequence {an }∞ n=1 inAsatisfyingan → rasn → ∞ andan > rforalln.Then,we have

Inparticular,wecanassumer

Proof. Itsufficestoprovethecaseofthesupremum.Itisobviousthat

Toprovepart(i),givenany ��> 0,accordingtotheconceptofsupremumsup

, thereexists a∗ ∈ A with a∗ ≥ r satisfying

)

Weconsiderthefollowingtwocases.

● Supposethat a∗ > r .Then,wehave

● Supposethat a∗ = r .Theassumptionsaysthatthereexistsasequence {an }∞ n=1 in A satisfying an ↓ a∗ as n → ∞ and an > r forall n.Since f isright-continuousand a∗ ∈ cl(A), wealsohave f (an ) → f (a∗ ) as n → ∞.Therefore,weobtain

Since �� canbeanypositivenumber,itfollowsthat sup {a∈A∶a≥r } f (a) ≤ sup {a∈A

Part(ii)canbesimilarlyobtained,andtheproofiscomplete.

Proposition1.2.9 Let {An }∞ n=1 and {Bn }∞ n=1 betwosequencesofsubsetsof ℝ satisfying An+1 ⊆ An andBn+1 ⊆ Bn foralln and

Then,wehave

Proof. Itisobviousthat

TheresultsfollowimmediatelyfromProposition1.2.3.

Proposition1.2.10 Let {An }∞ n=1 and {Bn }∞ n=1 betwosequencesofsetsin ℝ satisfying

An ⊆ An+1 andBn ⊆ Bn+1 foralln and ∞

Then,wehave

and

Proof. TheresultsfollowimmediatelyfromProposition1.2.4. ◾

Proposition1.2.11 Letfbeareal-valuedfunctiondefinedonasubsetAof ℝ,andletkbe aconstant.Then,wehave

Proof. Wehave

Anotherequalitycanbesimilarlyobtained.Thiscompletestheproof.

1.3Semi-Continuity

Let f ∶ ℝm → ℝ beareal-valuedfunctiondefinedon ℝm .Wesaythatthesupremum

supx∈S f (x ) is attained whenthereexists x ∗ ∈ S satisfying f (x ) ≤ f (x ∗ ) forall x ∈ S with x ≠ x ∗ .Equivalently,thesupremumsupx∈S f (x ) isattainedifandonlyif

sup x∈S f (x )= max x∈S f (x ).

Similarly,theinfimuminfx∈S f (x ) is attained whenthereexists x ∗ ∈ S satisfying f (x ) ≥ f (x ∗ ) forall x ∈ S with x ≠ x ∗ .Equivalently,theinfimuminfx∈S f (x ) isattainedifandonlyif

inf x∈S f (x )= min x∈S f (x )

Let �� =(x1 , … , xm ) beanelementin ℝm .RecallthattheEuclideannormof �� isgivenby ∥ �� ∥= √

Definition1.3.1 Let S beanonemptysetin ℝm

● Areal-valuedfunction f ∶ S → ℝ definedon S issaidtobe uppersemi-continuous at �� whenthefollowingconditionissatisfied:foreach ��> 0,thereexists ��> 0suchthat ∥ �� �� ∥<�� implies f (��) < f (��)+ �� forany �� ∈ S.

● Areal-valuedfunction f definedon S issaidtobe lowersemi-continuous at �� whenthe followingconditionissatisfied:foreach ��> 0,thereexists ��> 0suchthat ∥ �� �� ∥<�� implies f (��) < f (��)+ �� forany �� ∈ S.

Remark1.3.2 Wehavethefollowinginterestingobservations.

● If f isuppersemi-continuouon S,then f islowersemi-continuouson S.

● If f islowersemi-continuouon S,then f isuppersemi-continuouson S.

● Thereal-valuedfunction f iscontinuouson S ifandonlyifitisbothlowerandupper semi-continuousin S.

● If f isuppersemi-continuouson ℝ,then {�� ∶ f (��) ≥ �� } isaclosedsubsetof ℝm forall �� .

● If f islowersemi-continuouson ℝ,then {�� ∶ f (��) ≤ �� } isaclosedsubsetof ℝm forall �� .

Proposition1.3.3 Letf ∶ ℝm → ℝ beamulti-variablereal-valuedfunction,andleteach real-valuedfunctiongi ∶ ℝ → ℝ becontinuousatx0 ∈ ℝ fori = 1, , n.Then,thefollowing statementsholdtrue.

(i) Supposethatfislowersemi-continuousat ��0 ≡ (g1 (x0 ), … , gm (x0 )).Then,thecomposition functionh(x )= f (g1 (x ), … , gm (x )) islowersemi-continuousatx0 . (ii) Supposethatfisuppersemi-continuousat ��0 ≡ (g1 (x0 ), … , gm (x0 )).Then,thecompositionfunctionh(x )= f (g1 (x ), … , gm (x )) isuppersemi-continuousatx0 .

Proof. Toprovepart(i),since f islowersemi-continuousat ��0 ,givenany ��> 0,thereexists �� ∗ > 0suchthat ∥ �� ��0 ∥<�� ∗ implies f (��0 ) < f (��)+ ��.

Sinceeach gi iscontinuousat x0 for i = 1, … , n,given �� ∗ ∕√n,thereexists ��i > 0suchthat |x x0 | <��i implies | |gi (x )−

Let �� = min {��1 , … ,��m }.Then |x x0 | <�� impliesthattheinequality(1.13)issatisfiedfor all i = 1, … , n.Let �� ≡ (g1 (x ), … , gm (x )).Then

, whichimplies h(x0 )= f (g1 (x0 ), ,

), , gm (

))+ �� = h(

)+ ��, whichsaysthat h islowersemi-continuousat x0 .Part(ii)canbesimilarlyobtained.This completestheproof. ◾

Proposition1.3.4 Letf ∶ ℝm → ℝ beamulti-variablereal-valuedfunction,andleteach real-valuedfunctiongi ∶ ℝ → ℝ beleft-continuousatx0 ∈ ℝ fori = 1, , n.Then,thefollowingstatementsholdtrue.

(i) Assumethatthecompositionfunctionh(x )= f (g1 (x ), … , gm (x )) isincreasing.Iffislower semi-continuousat ��0 ≡ (g1 (x0 ), … , gm (x0 )),thenhislowersemi-continuousatx0 . (ii) Assumethatthecompositionfunctionh(x )= f (g1 (x ), … , gm (x )) isdecreasing.Iffisupper semi-continuousat ��0 ≡ (g1 (x0 ), … , gm (x0 )),thenhisuppersemi-continuousatx0 .

Proof. Toprovepart(i),since f islowersemi-continuousat ��0 ,givenany ��> 0,thereexists �� ∗ > 0suchthat

∥ �� ��0 ∥<�� ∗ implies f (��0 ) < f (��)+ ��.

Sinceeach gi isleft-continuousat x0 for i = 1, , n,given �� ∗ ∕√n,thereexists ��i > 0such that

0 < x0 x <��i implies | |gi (x )− gi (x0 )| |

√n for i = 1, … , n.

TheargumentintheproofofProposition1.3.3isstillvalidtoshowthatthereexists ��> 0 suchthat

0 < x0 x <�� implies h(x0 ) < h(x )+ ��.

For0 < x x0 <�� ,since h isincreasing,itfollowsthat h(x0 ) ≤ h(x ) < h(x )+ ��.

Therefore,weconcludethat

|x0 x | <�� implies h(x0 ) < h(x )+ ��,

whichsaysthat h islowersemi-continuousat x0

Toprovepart(ii),wecansimilarlyshowthatthereexists ��> 0suchthat

0 < x0 x <�� implies h(x ) < h(x0 )+ ��.

For0 < x x0 <�� ,since h isdecreasing,itfollowsthat h(x ) ≤ h(x0 ) < h(x0 )+ ��, whichsaysthat h isuppersemi-continuousat x0 .Thiscompletestheproof. ◾

Proposition1.3.5 Wehavethefollowingproperties.

(i) Supposethatthereal-valuedfunctionsf1 andf2 arelowersemi-continuousontheclosed interval [a, b].Then,theadditionf1 + f2 isalsolowersemi-continuousontheclosedinterval [a, b].

(ii) Supposethatthereal-valuedfunctionsg1 andg2 areuppersemi-continuousononthe closedinterval [a, b].Then,theadditiong1 + g2 isalsouppersemi-continuousonthe closedinterval [a, b].

Proof. Toprovepart(i),given ��> 0,thereexist ��1 ,��2 > 0suchthat

andthat

|

whichshowsthat f1 + f2 islowersemi-continuousat x0 Toprovepart(ii),given ��> 0,thereexist ��

,��2 > 0suchthat

|

Let �� = min {��

whichshowsthat

,wehave

2 isuppersemi-continuousat x0 .Thiscompletestheproof.

Proposition1.3.6 Wehavethefollowingproperties.

(i) Supposethatthereal-valuedfunctionsf1 andf2 arelowersemi-continuousontheclosed interval [a, b].Then,thereal-valuedfunctions min {f1 , f2 } and max {f1 , f2 } arealsolower semi-continuousontheclosedinterval [a, b]

(ii) Supposethatthereal-valuedfunctionsg1 andg2 areuppersemi-continuousononthe closedinterval [a, b].Then,thereal-valuedfunctions min {g1 , g2 } and max {g1 , g2 } are alsouppersemi-continuousontheclosedinterval [a, b]

Proof. Toprovepart(i),given ��> 0,thereexist ��1 ,��2 > 0suchthat |x x0 | <��1 implies f1 (x0 ) < f1 (x )+ ��,

andthat

|x x0 | <��2 implies f2 (x0 ) < f2 (x )+ ��.

Let �� = min {��1 ,��2 }.Then,for |x x0 | <�� ,wehave

{f1 (x0 ), f2 (x0 )} <

and

whichshowthatmin {f1 , f2 } andmax {f1 , f2 } arelowersemi-continuousat x0 .

Toprovepart(ii),given ��> 0,thereexist ��1 ,��2 > 0suchthat

|x x0 | <��1 implies

andthat

)+ ��>

|x x0 | <��2 implies g2 (x0 )+ ��> g2 (x ).

Let �� = min {��1 ,��2 }.Then,for |x x0 | <�� ,wehave

and

{g1 (x ), g2 (x )} , whichshowthatmin {g1 , g2 } andmax {g1 , g2 } areuppersemi-continuousat x0 .Thiscompletestheproof.

Proposition1.3.7 Wehavethefollowingproperties.

(i) SupposethatfisincreasingonasubsetDof ℝ.Thenfisleft-continuousonDifandonly iffislowersemi-continuousonD.

(ii) SupposethatgisdecreasingonasubsetDof ℝ.Thengisleft-continuousonDifandonly ifgisuppersemi-continuousonD.

Proof. Toprovepart(i),wefirstassumethat f isleft-continuousat x0 ∈ D.Then,given any ��> 0,thereexists ��> 0suchthat0 < x0 x <�� implies |f (x0 )− f (x )| <�� ,i.e. f (x0 ) < f (x )+ �� .For x0 ∈ D with0 < x x0 <�� ,since f isincreasing,wehave

f (x0 ) ≤ f (x ) < f (x )+ ��.

Therefore,weconcludethat |x0 x | <�� implies f (x0 ) < f (x )+ �� ,whichshowsthat f is lowersemi-continuousat x0 ∈ D

Conversely,weassumethat f islowersemi-continuousat x0 ∈ D.Then,givenany ��> 0, thereexists ��> 0suchthat |x0 x | <�� implies f (x0 ) < f (x )+ �� .If0 < x0 x <�� thenwe immediatelyhave f (x0 )− f (x ) <�� bythelowersemi-continuityat x0 .Since f isincreasing, wealsohave

f (x ) ≤ f (x0 ) < f (x0 )+ ��.

Therefore,weconcludethat0 < x0 x <�� implies |f (x0 )− f (x )| <�� ,whichshowsthat f is left-continuousat x0 ∈ D.

Toprovepart(ii),wefirstassumethat g isleft-continuousat x0 ∈ D.Then,givenany ��> 0,thereexists ��> 0suchthat0 < x0 x <�� implies |g(x0 )− g(x )| <�� ,i.e. g(x ) < g(x0 )+ �� .For x0 ∈ D with0 < x x0 <�� ,since g isdecreasing,wehave

g(x0 )+ �� ≥ g(x )+ ��> g(x ).

Therefore,weconcludethat |x0 x | <�� implies g(x ) < g(x0 )+ �� ,whichshowsthat g is uppersemi-continuousat x0 ∈ D.

Conversely,weassumethat f isuppersemi-continuousat x0 ∈ D.Then,givenany ��> 0, thereexists ��> 0suchthat |x0 x | <�� implies g(x ) < g(x0 )+ �� .If0 < x0 x <�� ,thenwe immediatelyhave g(x )− g(x0 ) <�� bytheuppersemi-continuityat x0 .Since g isdecreasing, wealsohave

g(x0 ) ≤ g(x ) < g(x )+ ��.

Therefore,weconcludethat0 < x0 x <�� implies |g(x0 )− g(x )| <�� ,whichshowsthat g is left-continuousat x0 ∈ D.Thiscompletestheproof. ◾

Let A beasubsetof ℝm .The characteristicfunction or indicatorfunction of A is definedby

A (x )= { 1for x ∈ A

Proposition1.3.8 LetSbeasubsetof ℝ,andlet �� L ∶ S → ℝ and �� U ∶ S → ℝ betwo boundedreal-valuedfunctionsdefinedonSsatisfying �� L (�� ) ≤ �� U (�� ) foreach �� ∈ S.Suppose that �� L islowersemi-continuousonS,andthat �� U isuppersemi-continuousonS.Let M�� =[�� L (�� ),�� U (�� )] for �� ∈ Sbeclosedintervals.Then,foranyfixedx ∈ ℝ,thefunction �� (�� )= �� ��M�� (x ) isuppersemi-continuousonS.

Proof. Foranyfixed ��0 ∈ S,wearegoingtoshowthat,givenany ��> 0,thereexists ��> 0 suchthat |�� ��0 | <�� implies �� (��0 )+ ��>�� (�� ).

Weconsiderthecasesof x ∈ M��0 and x ∉ M��0 .For x ∈ M��0 ,wehave �� (��0 )= ��0 .If |�� ��0 | < �� = �� ,wehave ��0 + ��>�� .Weconsiderthefollowingcases.

● Supposethat x ∉ M�� .Then,wehave �� (�� )= 0.Therefore,weobtain �� (��0 )+ �� = ��0 + ��> 0 = �� (�� ).

● Supposethat x ∈ M�� .Then,wehave �� (�� )= �� .Therefore,weobtain �� (��0 )+ �� = ��0 + ��>�� = �� (�� ).

Now,weconsiderthecaseof x ∉ M��0 ,i.e. x <�� L (��0 ) or x >�� U (��0 ).Inthiscase,wehave �� (��0 )= 0.

● For x <�� L (��0 ),let �� = �� L (��0 )− x .Since �� L islowersemi-continuousat ��0 ,thereexists ��> 0suchthat |�� ��0 | <�� implies �� L (��0 ) <�� L (�� )+ �� .Therefore,weobtain

● For x >�� U (��0 ),let �� = x �� U (��0 ).Since �� U isuppersemi-continuousat ��0 ,thereexists ��> 0suchthat |�� ��0 | <�� implies �� U (�� ) <�� U (��0 )+ �� .Therefore,weobtain

Thisalsosaysthat x ∉ M�� ,i.e. �� (�� )= 0for |�� ��0 | <��

Theabovetwocasesconcludethat �� (��0 )+ �� = ��> 0 = �� (�� ) for |�� ��0 | <�� .Thiscompletestheproof.

Proposition1.3.9 LetSbeasubsetof ℝ,andlet �� L ∶ S → ℝ and �� U ∶ S → ℝ betwo boundedreal-valuedfunctionsdefinedonSsatisfying �� L (�� ) ≤ �� U (�� ) foreach �� ∈ S.Suppose thatthefollowingconditionsaresatisfied.

● �� L isanincreasingfunctionand �� U isadecreasingfunctiononS. ● �� L and �� U areleft-continuousonS.

LetM�� =[�� L (�� ),�� U (�� )] for �� ∈ Sbeclosedintervals.Then,foranyfixedx ∈ ℝ,thefunction �� (�� )= �� ⋅ ��M�� (x ) isuppersemi-continuousonS.

Proof. TheresultfollowsimmediatelyfromPropositions1.3.8and1.3.7. ◾

Proposition1.3.10 LetSbeasubsetof ℝ.Foreachi = 1, , n,let �� L i ∶ S → ℝ and �� U i ∶ S → ℝ beboundedreal-valuedfunctionsdefinedonSsatisfying �� L i (�� ) ≤ �� U i (�� ) foreach �� ∈ S. Supposethatthefollowingconditionsaresatisfied.

● �� L i areincreasingfunctionand �� U i aredecreasingfunctiononSfori = 1, … , n.

● �� L i and �� U i areleft-continuousonSfori = 1, … , n.

LetM (i) �� =[�� L i (�� ),�� U i (�� )] for �� ∈ Sandfori = 1, … , nbeclosedintervals,andlet M�� = M (1) �� ×···× M (n) �� ⊂ ℝn

Givenanyfixed �� =(x1 , , xn )∈ ℝn ,thefunction �� (�� )= �� ⋅ ��M�� (��) isuppersemi-continuous onS.

Proof. Proposition1.3.9saysthatthefunctions ��i (�� )= �� ��M (i) �� (xi ) areuppersemicontinuouson S for i = 1, , n.For r ∈ S,wedefinethesets

Fr = {�� ∈ S ∶ �� (�� ) ≥ r } and F (i) r = {�� ∈ S ∶ ��i (�� ) ≥ r } for i = 1, , n

Theuppersemi-continuityof ��i saysthat F (i) r isaclosedsetfor i = 1, , n.Wewanttoclaim Fr = ⋂n i=1 F (i) r .Givenany �� ∈ Fr ,itfollowsthat �� ∈ M�� and �� ≥ r ,i.e. xi ∈ M (i) �� and �� ≥ r for i = 1, , n,whichalsoimplies ��i (�� ) ≥ r for i = 1, , n.Therefore,weobtaintheinclusion Fr ⊆ ⋂n i=1 F (i) r .Ontheotherhand,supposethat �� ∈ F (i) r for i = 1, , n.Itfollowsthat xi ∈ M (i) �� and �� ≥ r for i = 1, … , n,i.e. �� ∈ M�� and �� ≥ r .Therefore,weobtaintheequality Fr = ⋂n i=1 F (i) r ,whichalsosaysthat Fr isaclosedset,sinceeach F (i) r isaclosedsetfor i = 1, … , n. Therefore,weconcludethat �� isindeeduppersemi-continuouson S.Thiscompletesthe proof. ◾

Wesaythat S isadisjointunionofintervalsin ℝ when S canbeexpressedas S = ∞ ⋃ i=1 Ii

satisfying Ii ∩ Ij =∅ for i ≠ j,whereeach Ii isanintervalin ℝ

Proposition1.3.11 LetSbeadisjointunionofintervalsin ℝ,andlet �� L ∶ S → ℝ and �� U ∶

S → ℝ betwoboundedreal-valuedfunctionsdefinedonSsatisfying �� L (�� ) ≤ �� U (�� ) foreach �� ∈ S.For �� ∈ S,wedefinethefunctions

l(�� )= inf {x∈S∶x≥�� } �� L (x ) andu(�� )= sup {x∈S∶x≥�� } �� U (x ).

Thenlanduareleft-continuousonS.Moreover,lislowersemi-continuousonSanduisupper semi-continuousonS.

Proof. Given �� ∈ S,since S isadisjointunionofintervals,thereexistsasequence {��n }∞ n=1 in S satisfying ��n ↑ �� as n → ∞,whereweallow ��n = �� forsome n.Let An ={x ∈ S ∶ x ≥ ��n } and A ={x ∈ S ∶ x ≥ �� }

Thenitisobviousthat An+1 ⊆ An forall n and A ⊆ ⋂∞ n=1 An .For x ∈ ⋂∞ n=1 An ,itmeans x ∈ S and x ≥ ��n forall n.Bytakinglimit,weobtain x ≥ �� ,i.e. x ∈ A.Thisshowsthat A = ⋂∞ n=1 An .UsingProposition1.2.3,weobtain l(��n )= inf t∈An �� L (x ) → inf t∈A �� L (x )= l(�� ) for ��n ↑ ��.

Thissaysthat l isleft-continuouson S.Wecansimilarlyshowthat u isleft-continuous on S.Since l isdecreasingand u isincreasingon S,Proposition1.3.7saysthat l islower semi-continuouson S and u isuppersemi-continuouson S.Thiscompletestheproof. ◾

Let S beadisjointunionofintervalsin ℝ.Wewrite �� L (S) todenotethesetofallleft endpointsofsubintervalsin S,andwrite �� R (S) todenotethesetofallrightendpointsof subintervalsin S.Forany �� ∈ S∖�� R (S),i.e. �� ∈ S and �� ∉ �� R (S),itiscleartoseethatthere existsasequencein S satisfying ��n ↓ �� as n → ∞ with ��n >�� forall n.

Proposition1.3.12 LetSbeadisjointunionofclosedintervalsin ℝ,andlet �� L ∶ S → ℝ and �� U ∶ S → ℝ betwoboundedandright-continuousreal-valuedfunctionsdefinedonSsatisfying �� L (�� ) ≤ �� U (�� ) foreach �� ∈ S.LetM�� =[�� L (�� ),�� U (�� )] for �� ∈ Sbeclosedintervals.Then, thefunctions l(�� )= inf {x∈S

x≥�� } �� L (x ) andu(�� )= sup

(

) arecontinuousonS∖�� R (S).Moreover,for ��

S

�� R (S) and ��n ↓ �� asn → ∞ with ��n >�� for alln,wehavel(��

andu

Proof. AccordingtoProposition1.3.11,weremaintoshowthat l and u arerightcontinuouson S∖�� R (S).Wefirstnotethat S isaclosedset,i.e.cl(S)= S.Wearegoingto usepart(i)ofProposition1.2.8.Given �� ∈ S∖�� R (S),thereexistsasequence {��n }∞ n=1 in S satisfying ��n ↓ �� as n → ∞ with ��n >�� forall n.Let An ={x ∈ S ∶ x ≥ ��n } and A∗ ={x ∈ S ∶ x >�� }

L (x )= inf

Itiscleartoseethat An ⊆ An+1 forall n and ⋃∞ n=1 An ⊆ A∗ .For x ∈ A∗ ,i.e. x ∈ S and x >�� , since ��n ↓ �� ,thereexists ��n∗ satisfying �� ≤ ��n∗ < x ,whichsaysthat x ∈ ⋃∞ n=1 An .Therefore, weobtain ⋃∞ n=1 An = A∗ .UsingProposition1.2.4andpart(i)ofProposition1.2.8,for ��n ↓ �� with ��n >�� ,wehave l(��n )= inf x∈An �� L (x ) → inf x∈A

.

Therefore,weconcludethat l iscontinuouson S.Wecansimilarlyshowthat u iscontinuous on S.Since l isincreasingand u isdecreasing,wealsohave l(��n ) ↓ l(�� ) and u(��n ) ↑ u(�� ) as n → ∞ for ��n ↓ �� as n → ∞ with ��n >�� forall n,andtheproofiscomplete. ◾

Proposition1.3.13

LetSbeaclosedsubsetof ℝ,andlet �� L ∶ S → ℝ and �� U ∶ S → ℝ betwo boundedreal-valuedfunctionsdefinedonSsatisfying �� L (�� ) ≤ �� U (�� ) foreach �� ∈ S.Suppose that �� L islowersemi-continuousonS,andthat �� U isuppersemi-continuousonS.LetM�� = [�� L (�� ),�� U (�� )] for �� ∈ Sbeclosedintervals.Then,wehave

Proof. Since S isaclosedset,byProposition1.4.4(whichwillbegivenbelow),the semi-continuitiessaythattheimfimumandsupremumareattainedgivenby inf

wehave

thatis,

Toprovetheotherdirectionofinclusion,givenany x satisfying

wewanttoleadtoacontradictionbyassuming x

with ��

�� .Underthis assumption,sinceeach M�� isaboundedclosedinterval,itfollowsthat x

L (�� ) foreach �� ∈ S with �� ≥ �� or x >�� U (�� ) foreach �� ∈ S with �� ≥ �� .Sincetheinfimumandsupremum areattained,weobtain

whichcontradicts(1.16).Therefore,thereexists

completestheproof.

1.4Miscellaneous

Theconvexityoffuzzysetsisusuallyassumedforapplicationsinordertosimplifythediscussion.Theconceptofconvexsetin ℝn isgivenbelow.

Definition1.4.1 Let A beasubsetof ℝm .Wesaythat A is convex when,givenany x , y ∈ A,theconvexcombination ��x +(1 ��)y belongsto A forany0 <��< 1.

Definition1.4.2 Let f ∶ A → ℝ beareal-valuedfunctiondefinedonaconvexsubset A of ℝm .Thefunction f iscalled quasi-convex on A when,foreach x , y ∈ A,thefollowing inequalityissatisfied:

f (��x +(1 ��)y) ≤ max {f (x ), f (y)}

foreach0 <��< 1.

Itiswellknownthat f isquasi-convexon A ifandonlyiftheset

{x ∈ A ∶ f (x ) ≤ �� } isconvexforeach �� ∈ ℝ

Thefunction f iscalled quasi-concave on A when f isquasi-convexon A.Moreprecisely,thereal-valuedfunction f isquasi-concaveon A ifandonlyif

f (��x +(1 ��)y) ≥ min {f (x ), f (y)}

foreach0 <��< 1.Wealsoseethat f isquasi-concaveon A ifandonlyiftheset

{x ∈ A ∶ f (x ) ≥ �� }

isconvexforeach �� ∈ ℝ

Thefollowingwell-knownresultswillbeusedthroughoutthisbook.

Proposition1.4.3 (Apostol[3]) Wehavethefollowingresults

(i) Letfbeacontinuousreal-valuedfunctiondefinedonaconnectedsubsetSof ℝm .Suppose thatf (x ∗ ) < f (x ∘ ) forsomex ∗ , x ∘ ∈ S.Foreachysatisfyingf (x ∗ ) < y < f (x ∘ ),thereexists x ∈ Ssatisfyingf (x )= y.

(ii) Letfbeacontinuousreal-valuedfunctiondefinedonaboundedclosedintervalIin ℝ Supposethattherearetwopointsx , y ∈ Isatisfyingx < yandf (x ) ≠ f (y).Thenftakes everyvaluebetweenf (x ) andf (y) intheopeninterval (x , y)

(iii) Letf ∶ ℝp → ℝq beavector-valuedfunction.Supposethatfiscontinuousonaclosed andboundedsubsetXof ℝp .Thenf (X ) isaclosedandboundedsubsetof ℝq

(iv) Letf ∶ ℝp → ℝq beavector-valuedfunction.Supposethatfiscontinuousonaconnected subsetXof ℝp .Thenf (X ) isaconnectedsubsetof ℝq

Proposition1.4.4 (Royden[100]) Letfbeareal-valuedfunctiondefinedon ℝm ,andletK beaclosedandboundedsubsetof ℝm .Then,wehavethefollowingproperties.

(i) Supposethatfisuppersemi-continuous.Then,thesupremumisattainedinthefollowing sense sup x∈K f (x )= max x∈K f (x ).

(ii) Supposethatfislowersemi-continuous.Then,theinfimumisattainedinthefollowing sense inf x∈K f (x )= min x∈K f (x ).

Theorem1.4.5 (CantorIntersectionTheorem).Let {Q1 , Q2 , …} beacountablecollectionofnonemptysubsetsofatopologicalspace ℝm suchthatthefollowingconditionsare satisfied:

● Qk+1 ⊆ Qk fork = 1,2, …;

● eachQk isanonemptyboundedandclosedsubsetof ℝm forallk.

Then,theintersection ⋂∞ k=1 Qk isnonempty.

Proposition1.4.6 Let �� ∶[0,1]n → [0,1] beafunctiondefinedon [0,1]n .Supposethat

�� (��1 , ,��m ) ≥ �� ifandonlyif ��i ≥ �� foralli = 1, , n

Then,wehave �� (��1 , … ,��m ) = min {��1 , … ,��m } .

Proof. Since ��i ≥ min {��1 , … ,��m } forall i = 1, … , n,theassumptionsaysthat

�� (��1 , ,��m ) ≥ min {��1 , ,��m } bytaking �� = min {��1 , ,��m }.Ontheotherhand,supposethat ��(��1 , ,��m )= �� , i.e. ��(��1 , ,��m ) ≥ �� ,theassumptionsaysthat ��i ≥ �� forall i = 1, , n,whichimplies min {��1 , … ,��m } ≥ �� = �� (��1 , … ,��m ) .

Thiscompletestheproof. ◾

Proposition1.4.7 Let �� ∶[0,1]n → [0,1] beafunctiondefinedon [0,1]n .Supposethat �� (��1 , ,��m ) ≤ �� ifandonlyif ��i ≤ �� forsomei = 1, , n

Then,wehave �� (��1 , … ,��m ) = min {��1 , … ,��m } .

Proof. Supposethat ��i > min {��1 , … ,��m } forall i = 1, … , n.Thenmin {��1 , … ,��m } > min {��1 , … ,��m }.Thiscontradictionsaysthat ��i ≤ min {��1 , … ,��m } forsome i = 1, … , n.

Usingtheassumption,itfollowsthat �� (��1 , … ,��m ) ≤ min {��1 , … ,��m } bytaking �� = min {��1 , ,��m }.Ontheotherhand,supposethat ��(��1 , ,��m )= �� ,i.e. ��(��1 , ,��m ) ≤ �� ,theassumptionsaysthat ��i ≤ �� forsome i = 1, , n,whichimplies min {��1 , ,��m } ≤ �� = �� (��1 , ,��m )

Thiscompletestheproof. ◾

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.
[FREE PDF sample] Mathematical foundation of fuzzy sets 1st edition hsien-chung wu ebooks by Education Libraries - Issuu