Interval Analysis and its Applications to Optimization in  區(qū)間分析及其在優(yōu)化中的應(yīng)用_第1頁(yè)
Interval Analysis and its Applications to Optimization in  區(qū)間分析及其在優(yōu)化中的應(yīng)用_第2頁(yè)
Interval Analysis and its Applications to Optimization in  區(qū)間分析及其在優(yōu)化中的應(yīng)用_第3頁(yè)
Interval Analysis and its Applications to Optimization in  區(qū)間分析及其在優(yōu)化中的應(yīng)用_第4頁(yè)
Interval Analysis and its Applications to Optimization in  區(qū)間分析及其在優(yōu)化中的應(yīng)用_第5頁(yè)
已閱讀5頁(yè),還剩28頁(yè)未讀, 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說(shuō)明:本文檔由用戶(hù)提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡(jiǎn)介

PAGE

PAGE

1

IntervalAnalysisanditsApplicationstoOptimizationinBehaviouralEcology

by

JustinTung

CS490IndependentResearchReport

Instructor:DavidSchwartz

Date:December19,2001

TableofContents

Abstract 4

Introduction 5

IntervalAnalysis 5

BasicsandNotation 5

UncertaintyandApproximatingValues 5

IntervalArithmeticandFunctions 6

ForagingTheory 7

BasicsofForagingModels 7

SimplisticAnalyticForagingModel 8

TheOptimalResidenceTime 11

2.ResearchProblemandMethods 12

Motivation 12

ProblemswithFixedPointOptimizationinForagingModels 12

IntervalAnalysisasUncertaintyinMethod 13

ResearchProblem 13

Software 14

Methodology 14

Fixed-PointAnalysis 14

GeneralMethod 14

Algorithm:BisectionMethod 15

IntervalAnalysis 16

GeneralMethod 16

Algorithm:IntervalNewton’sMethod 16

VariationandConstraintsonParameters 17

3.NumericalAnalysisofModel 18

FixedPointAnalysis 18

GraphicalAnalysis 18

OptimizationandAnalysis 23

IntervalAnalysis 26

TrueSolutionsandIntervalOptimization 26

StabilityAnalysis 29

4.ConclusionsandFutureExploration 30

ResultsofNumericalStudy 30

ComparisonofFixedPointandIntervalRoots 30

ApplicationstoForagingModel 30

FutureExploration 32

Bibliography 33

Abstract

IntervalAnalysisisameansofrepresentinguncertaintybyreplacingsingle(fixed-point)valueswithintervals.Inthisproject,intervalanalysisisappliedtoaforagingmodelinbehaviouralecology.Themodeldescribesanindividualforaginginacollectionofcontinuouslyrenewingresourcepatches.Thismodelisusedtodeterminetheoptimalresidencetimeoftheforagerinaresourcepatchassumingtheforagerwantstomaximizeitsrateofresourceintake.Beforeapplyingintervalanalysis,fixed-point(non-interval)optimizationwillbedonetoserveasabasis.Certainparametersinthemodelwillthenbereplacedwithintervalsandinterval-basedoptimizationconducted.Acomparisonoftheintervalandfixed-pointresultswillbedoneaswellasanalysisofparameterintervalsandtheirconstraints,rootapproximations,andapplicationstothemodel.

Chapter1:Introduction

1.1IntervalAnalysis

1.1.1BasicsandNotation

ThispaperwillexplainonlythebasicsofIntervalAnalysis(IA)neededtounderstandthetopicscoveredandassumessomepriorknowledgeofIAandMatlab(see2.1.4regardingMatlab).ForaformalmathematicalintroductionandindepthcoverageofconceptsseeSchwartz(1999)orMoore(1966)listedinthereferences.Intervalanalysiswasinitiallydevelopedinthelate1960’stoboundcomputationalerroranditisadeterministicwayofrepresentinguncertaintyinvaluesbyreplacinganumberwitharangeofvalues(Schwartz17).Fixed-pointanalysisissimplyanalysisusingnon-intervalvalueswherethereisnouncertaintyinthevalues.Asaresult,IAuncertaintyconceptscanbeusedtomodelvaryingbiologicalparametersintheecologicalmodeltobeexploredinsection1.2andalsotoframefixed-pointresults.

IA’smathematicaldefinitionsandnotationsareextendedfromsettheoryandorderednumericalsetscalledintervals(Schwartz30).Thispaperconsidersclosedintervalanalysiswiththefollowingdefinitionsofaninterval(usingMatlabupperbound,lowerboundstylenotation):

inf(x)–denotestheinfinum,orlowerboundofx

sup(x)–denotesthesupremumorupperboundofx

1.1.2UncertaintyandApproximatingValues

Thereareaseveralusefulquantitiesrelatedtotheconceptoftheinterval:size,radius,andmidpoint.Thesize(orthickness)ofanintervalindicatestheuncertaintyinavalueandisspecifiedasawidth:w(x)=sup(x)-inf(x)(Schwartz32-33).Intervalswithzerothicknessarecrispintervalswhereasnon-crispintervalssaidtobethick.Theconceptsofradiusandmidpointareusefulindescribingintervalsaswellasconstructingthem.Theradiusandmidpointaredefinedas(Schwartz33):

rad(x)=w(x)/2

mid(x)=(sup(x)+inf(x))/2

Toconstructanewinterval,onewayistouseanoriginalvalue,whichisavaluethatsuppliesthemidpointpointofanewinterval.Then,acertainradius(uncertainty)canbeaddedtoandsubtractedfromtheoriginalvaluetoobtainanewinterval(Schwartz35).Similarly,themidpointcanalsoserveasanapproximationtoavaluewithanerrorofplusorminustheradius.Usingthesedefinitions,thepercentageuncertaintyinamidpointvaluewouldbe:p=100*rad(x)/mid(x)(Schwartz148).

1.1.3IntervalArithmeticandFunctions

Theresultsandpropertiesofintervalarithmeticwillbeomittedforthissection;however,IrecommendreferringtoSchwartz(1999)tounderstandthebasicsofintervalarithmetic.Thefundamentalprinciplesinintervaloperationsareindependenceandextremes.Independencemeansnumericalvaluesvaryindependentlybetweenintervalsandextremesmeansintervaloperationsgeneratethewidestpossibleboundsgiventherangesofvalues(Schwartz37-38).Interval-valuedfunctionsfollowfromintervalarithmeticofwhichtherearetwotypes:intervalextensionsandunitedextensions(ortruesolutionsets).Intervalextensionsarefunctionswhereintervalarithmeticisappliedtocalculateresults.Unitedextensionsaremorecomputationallyintensiveandinvolvecalculatingfixed-pointresultswithallpossiblecombinationsofvariableintervalendpoints.Thedisadvantageoftheintervalextensionsisthattheycanoverexpandthetruesolutionsetsofafunction(Schwartz45-49).Thisqualityofintervalextensionsisunfortunatesincebothtypesofextensionsguaranteecontainmentofallpossiblenumericalresultsofthefunctiongiventheinputs.Also,bothextensionssatisfyapropertycalledinclusionmonotonicity(giveninputs,anextensiongeneratesthewidestpossiblebounds),whichissimilartotheextremesprincipleofintervalarithmetic(Schwartz56).

1.2ForagingTheory

1.2.1BasicsofForagingModels

Foragingmodelsingeneralstudytwobasicproblemsofaforager:whichfood/preyitemstoconsumeandwhentoleaveanareacontainingfood(aresourcepatch).Thispaperwillconcentrateonthelatterasanoptimizationproblem.Beforegoingintodetailsofthemodel,itisimportanttounderstandtheframeworkofforagingmodels.StephensandKrebspointoutthatforagingandoptimalitymodelshavethreemaincomponents,decision,currency,andconstraintassumptions(5).Decisionassumptionsdeterminewhichproblems(orchoices)oftheforageraretobeanalyzedandthesechoicesareusuallyexpressedasvariables.Theoptimizationproblemcomesfromassumingbehaviourandevolutionarymechanismsoptimizetheoutcomeofaforager’schoices.Currencyassumptionsprovidemeansofevaluatingchoices.Thesechoicesusuallyinvolvemaximization,minimization,orstabilityofasituation.Choiceevaluationisembodiedinthecurrencyfunction(arealvaluedfunction),whichtakesthedecisionvariablesandevaluatestheiroutcomeintoasinglevalue.Constraintassumptionsarelimitationstothemodelandrelatedecisionvariableswiththecurrency.Limitationscanbegeneralizedto2types,extrinsic(environmentlimitsonanimal)andintrinsic(animal’sownlimitations).Also,therearethreegeneralconstraintassumptions(alsoassumedbythemodelinsection1.2.2)forconventionalforagingmodels:

1)Exclusivityofsearchandexploitation–thepredatorcanonlyconsumeorsearchforpatches/preyandnotperformbothactionsarethesametime

2)SequentialPoissonencounters–items(preyorpatches)areencounteredoneatatimeandthereisaconstantprobabilityregardingprey/patchmeetingsinashorttimeperiod

3)Completeinformation–theforagerbehavesasifitknowstherulesofthemodel

Thesethreeconceptsofdecision,currency,andconstraintprovidemeansofoptimizationgivenchoices,howtodeterminetheirsuccess,andlimitations(StephensandKrebs6-11).

1.2.2SimplisticAnalyticForagingModel

Giventhestructureofaforagingmodel,itiseasytoframeamodelexaminingtheforagingofasingleanimaloveracollectionofdistinctresourcepatches.Ihavetakenthemodelalongwithitsdecision,currency,andconstraintassumptionsfromWilson(2000)soderivationsofthemodel’sequations,itsorigins,andananalysisandextensionsofthemodelinCcanbefoundinhisbook.Therulesoftheforagerinthemodelarethattheanimalstaysforafixedtimebeforemovingtoanewresourcepatch,timeisdiscrete,andpatchresourcevalues(biomass,energy,etc)growlogistically.Thedecisionassumptionliesinthedeterminationofthefixedtimevalue.Thecurrencyfunctionallowsustooptimizetheanimal’ssituationgiventhisfixedtimeandalsoallowsustoapplyconstraintstothemodel(Wilson152).Despiteitsname,thesimplisticanalyticforagingmodelactuallycharacterizesforagingsimulationresultswellusingthemodel’sspecifications.Hereisalistofparameterstakeninbythemodel:

Toderivethemodel,wecanstartanalyzingtheresourcesideassumingthattheithpatchwithouttheforagerconsuminggrowslogistically:

Wheretistimeandriistheamountofresourcesintheithpatch.Ifaforagerentersagivenpatchf,resourcedynamicscanbemodeledasfollows:

Inthefthpatch,theconsumerdecreasestherateofgrowthbyafactorrelatedtobeta,theconsumingrate.Tomodeloverallpatchgrowth,averagepatchgrowthforN-1identicalpatchesisaddedtothepatchgrowth(ordecay)ofthefthpatch:

Toachieveanequilibriumresourcedensityr*,wesetdr/dt=0andsolveforryielding:

AquickanalysisofthelimitasNapproachesinfinityshowsthattheforager’seffectisinsignificantatequilibriumsincethetermcontainingbetagoestozeroandr*goestoKasexpected(Wilson153-154).Theresourcesideprovidesuswithanenvironmentandextrinsicconstraintsthatwillaffectthecurrencyfunctionwhichliesontheforagersideofthemodel.Keytoresourceexploitationmodelsisthegainfunction,g(t).Thegainfunctionspecifiestheamountconsumedfromaresourcepatchgiventimet.Assumingtheforagerlandsonaresourcepatchalwaysinequilibriumr*,g(t)isthetimeintegralofitsinstantaneousconsumptionrateminusitsmetaboliccosts.Metaboliccostsrepresentintrinsicconstraintssincetheanimalmust“pay”thesecostswhenforaging.

Thextandxmconstraintsactasintrinsiclimitationsonthemodelsincethetravelingcostspreventstheforagerfrommovingquicklyfrompatchtopatchandskimmingresources,whilethemetaboliccostcausestheforagertogatherresourcesforthreatofdeath.Inordertoevaluateg(t),werequireananalyticalsolutiontorf,whichmeasurestheresourcesinthepatchtheforagerisin.Solvingthefirstorderdifferentialequationfromtheresourcesidederivationsforrfusingseparationofvariablesyields:

Thensubstitutingthisequationintog(t)andsolvingtheintegralweget:

Usingthisgainfunction,thecrucialequationfromtheforagerperspective,thenetforagingratefunctioncanbecalculatedas:

R(t)isthecurrencyfunctionforourmodelsinceitisthebasisofchoiceevaluationandoptimizationforthemodel(Wilson154-155).Italsocombinesthedecisionvariablesandconstraintsintoonevalueandwillbeabasisforgraphicalanalysislateron.

1.2.3TheOptimalResidenceTime

Assumingbehaviouralandevolutionarymechanismsdriveforagerstooptimizethetimespentoneachpatch.Thisassumptionimpliesthattheywillstaylongenoughtooptimizetherateofresourceconsumption,r(t).Mathematically,thischoiceimpliesthemaximizationofr(t):

wheret*iscalledtheoptimalresidencetime.Forratemaximization,r’’(t*)<0mustalsobechecked;however,assumingr(t*)isatamaximum,weobtaintheequation:

whichrelatesr(t)tothederivativeofthegainfunction.Thesecondfunctionaboveisthefunctiontobeusedforrootfinding.Essentially,theoptimaltimetoleaveapatchiswhentheexpectedrateofresourcereturndecreasestotheaveragerateofreturnofanewpatch.Thisresult,whichisastatementofthemarginalvaluetheoremforforaging,isareasonableestimationofanimalbehaviourconsideringaforager’sdesiretomaximizeonitsresourceintake(Wilson156-157).Oneprobleminoptimizationproblemsisfiguringoutwhetheranoptimalpointexistsandinthiscase,wemustbesurer’’(t*)<0andbesuresuchat*exists.Unfortunately,duetothecomplexityofg(t)isitnotpossibletosolveforanexplicitsolutionoft*sotojustifyexistenceofasolution,weturntotheoreticaljustificationsand,laterinsection3.1.1,graphicalmeans.Duetotheconditionsspecifiedinforagingmodels,gainfunctionsare“well-defined,continuous,deterministic,andnegativelyaccelerated”functions(StephensandKrebs25-26).Thisoutcomeresultsfromassumingpatchresourcesaresufficient(i.e.theequilibriumresourceissufficientlylarge)enoughthat,whentheforagerentersapatch,r(t)willreachamaximumandthendecreaseuntilthegainfunctionreachesanasymptoticmaximumwhenfurthertimespentintheresourcepatchdoesnotyieldsignificantlymoregaininresources.Thereasonforthegainfunction’spropertiesistheassumptionthatpatchescontainafinitenumberofresourcesandforagingdepletesthem.Thisassumption;however,reliesontheassumptionsaboutresidencetime,foragingrates,andresourcepatchdynamicsingeneral(StephensandKrebs25-26).Duetothesimplenatureofthemodel,thesegainfunctionpropertiesholdsor(t)doesreachamaximumatt*andr’’(*t)<0.

Chapter2:ResearchProblemsandMethods

2.1Motivation

2.1.1ProblemswithFixedPointOptimizationinForagingModels

StephensandKrebs(1986)discussvariouslimitationsandcriticismsofbehaviouralecologyoptimalitymodels.Onecriticismhastodowith“staticversusdynamic”modelingsincebasicforagingmodelsoftendonottaketheanimal’sstateintoaccount(i.e.whetherananimalisstarvingorfullyrestedandfed)(StephensandKrebs34).Also,aproblemthatoccursduringtestingphasesofamodeliswhenitbreaksdown.Atthatpoint,theecologistmustre-analyzethemodeltofindwhatisincorrect,oftencheckingconstraints(StephensandKrebs208)

2.1.2IntervalAnalysisasanUncertaintyMethod

IAcanaddressbutnotcompletelysolvetheproblemsstatedabove.OneofthestrengthsofIAisitsabilitytoevaluateawholerangeofvaluesinonecalculationthatwouldtakeaninfinitenumberoffixed-pointcalculationstoproduce.Asaresult,IAcouldsimulatethepresenceofmultiplestatesofananimaland/oritsenvironmentbyplacinguncertaintyinthemodel’sparameters.Thismethodprovideseasydeterministicimplementationofmultiplestatemodelsandproducesrangesofvaluesforevaluation.Thismethodpartiallyaddressesthesecondproblemofwhenamodelbreaksdown.Amodelcouldbreakdownduetoincorrectassumptionsaboutconstantforagerorenvironmentstates.AnotherapplicationofIAtotestingisthatIAisnumericallysuperiorwhenitcomestotestingdifferentacceptableuncertaintiesinvaluescouldhelpidentifyproblemsinthemodelorunrealisticassumptions.

2.1.3ResearchProblem

ThepurposeofthispaperistointroduceIAmethodstothesimplisticanalyticforagingmodelandcalculateintervaloptimalresidencetimesforintervalparameters.Atsametime,solvingthefixedpointoptimalresidencetimewillprovideframeworkfromwhichtoanalyzetheintervalresults.TheoptimizationwillbedoneforpatchsizesN=3,5,10,and20.Theparameterswithuncertaintywillbe:

Theseuncertaintiescouldbestrengthenedwithfieldworkstudies;however,forsimplicitytheyaredeterminedapriori.Aftercomputingintervaloptimalresidencetimes,theintervalsandtheirfixed-pointapproximationswillbecomparedtothefixed-pointoptimaltimestocomparealgorithmsandtoanalyzethefunctions’behaviourunderbothmethods.Onamoretheoreticalside,stabilityanalysisofthemodelwillbeconducted.Thisanalysisinvolvesvaryingoneuncertainparameter,whileholdingtheothersconstantuntilthemodelfails.Therefore,stabilityanalysisisusedtoseeperformanceofthemodelunderparameteruncertaintyperturbations.Conditionsforfailurewillbespecifiedinsection.

2.1.4Software

ThelanguagetobeusedisMatlabversion5.3withanadd-ontoolboxcalledINTLABprogrammedbySiegfriedRump(2001).RefertothereferencesfordocumentationontheINTLABtoolbox.Forthispaper,theINTLABtoolboxisusedtoprovideintervaldatastructures,implementationofintervalarithmeticandinterval-valuedfunctions,aswellasbasicfunctionsforradius,midpoint,andintersectionintervalfunctionsinMatlab.

2.2Methodology

2.2.1Fixed-PointAnalysis

GeneralMethod

Rootfunc(t)specifiedbelowisthefunctionwhoserootweareseeking:

Sincerootfunc(t)hasrelativelycheapfunctionevaluationsitisusefultoperforma“graphicalsearch”fortheroot(VanLoan294).Thisprocedureinvolvesplottingthefunctioninthetimeintervalofinterestandexaminingitsroots.Inadditiontothisfunction,duringthefixed-pointanalysis,wewillalsoplotr(t)tosearchfortheexistenceofmaximumsaswellasrootfunc’’(t)toconfirmthatrootfunc’’(t)isindeednegativeintheintervalofinterest.Although,graphicalsearchesarerathertrivial,theyprovidealargeamountofinformationconfirmingtheoreticalconclusionsinsection1.2aswellasenablingapictorialviewoftheobjectiveandrelatedfunctions.Anotheruseoftheplottingoffunctionsbeforeoptimizationistousetheplotstogeneratestartingintervalsforiterationsofrootfindingmethods.Inordertoplotrootfunc,r(t),andr’’(t)itisnecessarytoimplementequationsforg(t),g’(t),andr’’(t).Thedetailsforcalculatingg’(t)andr’’(t)areleftoutsinceg(t)isarathermessyfunction,butthederivativesareimplementedintheMatlabcodeforsection3.1.1.Assumingthepropertiesofthegainfunctiondiscussedin1.2.3,whichwillbeconfirmedinsection3.1.1,itisnecessarytochooseanalgorithmthatwillproducearootgiventheconditionsofrootfunc(t).

Algorithm:BisectionMethod

Sincerootfunc(t)iscontinuousandchangessignwithintheintervalofinterest,thebisectionmethodcanbeused.Thismethodinvolvescalculatingasequenceofsmallerandsmallerintervalsthatbracket(contain)therootofrootfunc(t).Themainalgorithmproceedsasfollows(ifrootfunct(t)=f(t))givenabracketinginterval[a,b]:

assumef(a)f(b)0andletm=(a+b)/2

eitherf(a)f(m)0orf(m)f(b)0

inthefirstcaseweknowtherootisin[a,m]elseitisit[m,b]

Ineithersituation,thesearchintervalishalvedandthisprocesscanbecontinueduntilasmallenoughintervalisobtained.Sincethesearchintervalishalvedwitheachiteration,thebisectionmethodexhibitsO(n)convergence.Theonlytrickypointsaretooptimizethemethodsothatonlyonefunctionevaluationisrequiredperiterationafterthefirstandtoestablishasafeconvergencecriterionsothatthetoleranceintervalisnotsmallerthanthegapinthefloatingpointnumbersbetweenaandb.VanLoanprovidesthecoreofthecodeforthebisectionmethodwithslightmodificationstofittheparametersofthemodel(280).AlthoughthebisectionmethoddoesnotexperienceO(n^2)convergenceliketheNewtonmethod,rootfunc(t)issimpleenoughthatitconvergesquicklyforpracticalpurposes.Also,itissimple(algorithmicallyanddoesn’trequirerootfunc’(t)implementation)andtranslateswellintotheideaofsearchingusingastartinginterval[a,b],whichwewilluselaterintheintervalrootfinding.

2.2.2IntervalAnalysis

GeneralMethod

Whenchangingfromfixed-pointtointervalbasedrootfindingtherearesomeimmediatedifferences.Therootisnolongeracrispintervalsinceiterationsusinganinterval-valuedfunctionproduceintervals.Asaresult,convergencecriterionsandthegeneralmethodsofrootfindingmusthaveasetvaluedapproach.Becausetheoptimalresidencetimeswillinherentlybethickintervalssincetherootfunc(t)isnowanintervalfunctionwithuncertainparameters,convergencewillbesolvedsimplebysettingamaximumnumberofiterations.ThereasonthisconvergenceguaranteesanenclosureoftherootisduetotheIntervalNewtonmethod,whichisbasedonthefixed-pointone.

IntervalNewtonMethod

Fordetailsofthemathandconvergencepropertiesofthealgorithm,refertoKulischetal.(2001).Thisalgorithm,whenfindingrootsoffixed-pointfunctionsexhibitsO(n^2)convergence.LiketheBisectionMethod,itrequiresabracketingintervaltobeginandwitheachiterationgeneratessmallerandsmallerintervals(ifpossible),whichareboundedbyintersectionswithpreviousiterations.Thealgorithmisasfollows:

wherethex’sareintervals,m(x)isthemidpointofatheintervalx,andfisthefunctionwhoserootweseek(Kulischetal.35-36).Thesimilarityofthisalgorithmtothefixed-pointNewtonmethodisthatastartingintervalmustbesuppliedandtheintervalsizeisdecreasedusingaf(x)/f’(x)termduringiterations.Thisalgorithmisalmostassimpleasthebisectionmethodsinceaneasyconvergencecriterionhasbeenspecified;however,theintervalNewtonrequiresimplementationofrootfunc’(t)andrequiresforanevaluatedintervalx.Throughcomputationaltrials,Ihavedecidedtoincreasethecomplexityofthefunctionevaluationsforthef(x)/f’(x)bycomputingitsunitedextensioninsteadofaintervalextension.Sincebothf(x)andf’(x)involvemanyintervalarithmeticcalculationsinanintervalextension,thevaluesareoverexpandedfromtheirtruesolutionsetandincomputingoptimalresidencetimesweareinterestedinfindingtightboundsontherootgivenmodeluncertainties.Asaresult,theintervalNewtonstepiscalculatedbyfindingtheminimumsandmaximumsoff(x)andf’(x)givenallthecombinationsoftheendpointsoftheparametersandthetimeintervalandproducingunitedextensionvaluesforbothquantities.

VariationandConstraintsonParameters

Asoutlinedin2.1.3,percentuncertaintiesinthealpha,beta,xt,andxmparameterswillbeaddedwhenanalyzingthemodelusingintervalrootfinding.Thisvariationofparametersservestoaddresstheproblemswithfixed-pointoptimizationoutlinedin2.1.1.Thepotentialuseforintervalparametersliesinmodeltesting,simulatingarangeofenvironmentandforagerstates,andrelaxationofconstraints.Besidesreplacingparameterswithintervals,wealsowanttoconductthestabilityanalysismentionedin2.1.3.Thisprocessinvolvesincreasinguncertaintiesofagivenparameterwhilekeepingallotherparametersconstantuntilthemodelfails.AfterspecifyingtheintervalNewtonalgorithmwecanconstructafailurecondition.FailureofthemodelcanbeconsideredtooccurwhentheintervalNewtonmethodreturnsoftheinitialintervalsuppliedtotheintervalrootfinderastheoptimaltimeintervalforanyNnumberofpatches(i.e.thealgorithmwasunabletoprovidetighterboundsforoptimalresidencetimethantheinitialinterval).Thisinitialintervalwillbechosentobeasufficientlythickintervalfromfixed-pointanalysis,whichenclosesthetimeintervalofinterest.Asaresult,theintervalwillbetheonechosentoinitiatethebisectionmethod.Thereareotherconditionsthatcouldclassifyfailure;however,fromamodelstandpoint,averythickintervalforanoptimalresidencetimeisnotveryusefulsinceitimpliestoomuchvariabilityinaforager’sbehaviour.Consequently,stabilityanalysisisanumericallyintensiveprocedureinvolvinggradualincreaseduncertaintiesofaparameteruntilmodelfailure.Itisinterestingmorefromatheoreticalstandpointsinceitdescribeslimitationsofthemodelaswellasextremesituationsandtheireffectsonaforager’soptimalresidencetime.

Chapter3:NumericalAnalysisofModel

Inthischapter,IwillincludeMatlabcodeofkeyfunctionsimplementingthegeneralmethodsdiscussedin2.2,supportingfunctions,andalgorithmsastheyareusedinthenumericalanalysisofthemodel.

3.1Fixed-PointAnalysis

3.1.1GraphicalAnalysis

Insection,thegraphsofr(t),rootfunc(t)androotfunc’(t)wereseentobeinformativeforselectinganinitialintervaltobeginthebisectionandintervalNewtonmethods.Also,thesegraphshelpvisualizethebehaviourofthemodelastimepassesaswellasconfirminganoptimalresidencetimeexists.Thefollowingfunctionsarerequiredtoimplementthefunctionsandgraphthem.Also,thefollowingparameterspecificationsfromWilsonareused(155):

functiongain=gain(t,alpha,K,beta,xt,xm,N)

%GAIN(t,alpha,K,beta,xt,xm,N)GainFunction

%

%Generalevaluationoftheg(t)functionwhich

%takesinthevariablesinorder:time,patchgrowth

%rate,patchcarryingcapacity,consumingrate,

%travelcost,metabolicrate,andnumberof

%resourcepatches

rstar=(1-beta/alpha/N)*K;

deltaBA=beta-alpha;

num=(alpha*rstar+...

K*(deltaBA))*exp(deltaBA*t)-...

alpha*rstar;

denom=K*deltaBA;

inLog=num/denom;

gain=(beta*K/alpha)*(log(inLog)-deltaBA*t)-...

(xt+xm*t);

functionintakeRate=r(t,alpha,K,beta,xt,xm,N)

%R(t,alpha,K,beta,xt,xm,N)IntakeRateFunction

%

%Calculatesthenetintakerate.Takesinthevariables

%inorder:time,patchgrowthrate,patchcarrying

%capacity,consumingrate,travelcost,metabolicrate,

%andnumberofresourcepatches

g=gain(t,alpha,K,beta,xt,xm,N);

intakeRate=g./t;

functiongainprime=gainprime(t,alpha,K,beta,xt,xm,N)

%GAINP(t,alpha,K,beta,xt,xm,N)GainFunctionDerivative

%

%Generalevaluationoftheg'(t)function

%

%Takesinthevariablesinorder:time,patchgrowth

%rate,patchcarryingcapacity,consumingrate,

%travelcost,metabolicrate,andnumberof

%resourcepatches

rstar=(1-beta/alpha/N)*K;

deltaBA=beta-alpha;

exppart=(alpha*rstar+K*deltaBA)*exp(deltaBA*t);

num=deltaBA*exppart;

denom=exppart-alpha*rstar;

gainprime=(beta*K/alpha)*(num./denom-deltaBA)-xm;

functionrootfunc=rootfunc(t,alpha,K,beta,xt,xm,N)

%ROOTFUNC(t,alpha,K,beta,xt,xm,N)RootFunction

%

%Thisfunctiontheonewewanttofindtherootof

%Takesinthevariablesinor

溫馨提示

  • 1. 本站所有資源如無(wú)特殊說(shuō)明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶(hù)所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁(yè)內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒(méi)有圖紙預(yù)覽就沒(méi)有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫(kù)網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶(hù)上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶(hù)上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶(hù)因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。

最新文檔

評(píng)論

0/150

提交評(píng)論