ebook img

Advances in Metaheuristics for Hard Optimization PDF

483 Pages·2008·12.971 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Advances in Metaheuristics for Hard Optimization

Natural Computing Series SeriesEditors: G.Rozenberg Th.Bäck A.E.Eiben J.N.Kok H.P.Spaink LeidenCenterforNaturalComputing AdvisoryBoard: S.Amari G.Brassard K.A.DeJong C.C.A.M.Gielen T.Head L.Kari L.Landweber T.Martinetz Z.Michalewicz M.C.Mozer E.Oja G.Pa˘un J.Reif H.Rubin A.Salomaa M.Schoenauer H.-P.Schwefel C.Torras D.Whitley E.Winfree J.M.Zurada Patrick Siarry · Zbigniew Michalewicz (Eds.) Advances in Metaheuristics for Hard Optimization With167Figuresand82Tables 123 Editors SeriesEditors PatrickSiarry G.Rozenberg(ManagingEditor) UniversityofParis12 [email protected] LaboratoryLiSSi 61AvenueduGénéraldeGaulle Th.Bäck,J.N.Kok,H.P.Spaink 94010Créteil,France LeidenCenterforNaturalComputing [email protected] LeidenUniversity NielsBohrweg1 ZbigniewMichalewicz 2333CALeiden,TheNetherlands SchoolofComputerScience UniversityofAdelaide A.E.Eiben Adelaide,SA5005,Australia VrijeUniversiteitAmsterdam [email protected] TheNetherlands LibraryofCongressControlNumber:2007929485 ACMComputingClassification(1998):F.2,G.1,I.2,J.6 ISSN 1619-7127 ISBN 978-3-540-72959-4 SpringerBerlinHeidelbergNewYork Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialisconcerned,specif- icallytherightsoftranslation,reprinting,reuseofillustrations,recitation,broadcasting,reproductiononmicrofilmor inanyotherway,andstorageindatabanks.Duplicationofthispublicationorpartsthereofispermittedonlyunderthe provisionsoftheGermanCopyrightLawofSeptember9,1965,initscurrentversion,andpermissionforusemustalways beobtainedfromSpringer.ViolationsareliableforprosecutionundertheGermanCopyrightLaw. SpringerisapartofSpringerScience+BusinessMedia springer.com ©Springer-VerlagBerlinHeidelberg2008 Theuseofgeneraldescriptivenames,registerednames,trademarks,etc.inthispublicationdoesnotimply,eveninthe absenceofaspecificstatement,thatsuchnamesareexemptfromtherelevantprotectivelawsandregulationsandtherefore freeforgeneraluse. CoverDesign:KünkelLopka,Werbeagentur,Heidelberg TypesettingandProduction:LE-TEXJelonek,Schmidt&VöcklerGbR,Leipzig Printedonacid-freepaper 45/3180/YL 543210 Preface The community of researchers claiming the relevance of their work to the field of metaheuristics is growing faster and faster,despitethe factthat the term itself has notbeenpreciselydefined.Numerousbookshavebeenpublishedspecializinginany oneofthemostwidelyknownmethods,namely,simulated annealing, tabusearch, evolutionaryalgorithms,antcolonyalgorithms,particleswarmoptimization,butat- temptstobringmetaheuristicsclosertogetherarescarce.Yetsomecommonfeatures clearlyappearinmostmetaheuristics, suchastheuseofdiversification toforcethe exploration of regions of the search space, rarely visited until now, and the use of intensification,toinvestigatethoroughlysomepromisingregions.Anothercommon featureistheuseofmemorytoarchivethebestsolutionsencountered.Onecommon shortcomingofmostmetaheuristicsisthedelicatetuningofnumerousparameters; thetheoreticalresultsavailablearenotsufficienttohelptheuserfacinganew,difficult optimizationproblem. Thegoalofthisbookistocollectstate-of-the-artcontributions thatdiscussre- centdevelopmentsinaparticularmetaheuristicorhighlightsomegeneralideasthat proved effective in adapting a metaheuristic to a specific problem. Some chapters are overview-oriented while others describe recent advances in one method or its adaptationtoareal-worldapplication.Thebookconsistsofchapterscoveringtop- icsfromvariousareasofmetaheuristics,includingsimulatedannealing(chapters), tabusearch(chapters),antcolonyalgorithms(chapters),generalpurposestudies onevolutionary algorithms (chapters),applications ofevolutionary algorithms ( chapters),andmiscellaneousmetaheuristics(chapters). The first chapter on simulated annealing, by Chandra Sekhar Pedamallu and LinetÖzdamar,isdevotedtoacomparisonofasimulatedannealing(SA)algorithm, anintervalpartitioning(IP)algorithm,andahybridalgorithmintegratingSAintoIP. Allthree methods, developed forsolving the continuous constrained optimization problem, are equipped with a local solver that helps to identify feasible stationary points.Theperformancesaretestedonasuiteofbenchmarkproblemscollected fromdifferentsourcesintheliterature. Inthesecond chapter,HoracioMartínez-Alfaro appliessimulated annealing to linkagesynthesisofafour-barmechanismforagivennumberofdesiredpathpoints. VI Preface Severalexamplesareshowntodemonstratethatapathcanbebetterspecified,since theuserisabletoprovidemoreprescribedpointsthantheusuallimitednumberof fiveallowedbytheclassicalmethods. ThefollowingchapterbyRicardoP.Beausoleildealswithnonlinearmultiobjec- tiveoptimization.ThechapterintroducesanewversionoftheMultiobjectiveScatter Search(MOSS)algorithm,applyingamulti-starttabusearchandconvexcombina- tionmethodsasadiversification generation method.Aconstraint-handling mech- anism isincorporated todeal with constrained problems. Theperformanceof this approachistestedthroughtestproblems. Muhammad A.Tahir andJamesE.Smiththenproposeanewapproachtoim- provingtheperformanceofthenearestneighbor(NN)classifier.Thetechniquecom- bines multiple NN classifiers, where each classifier uses a different distance func- tionandpotentiallyadifferentsetoffeatures,determinedthroughacombinationof tabusearchandsimplelocalneighborhoodsearch.Comparisonwithdifferentwell- knownclassifiersisperformedusingbenchmarkdatasetsavailableintheliterature. ThefollowingchapterbyAdemKalinliandFatihSarikocpresentsanewparal- lelantcolonyoptimizationalgorithmaimedatsolvingcontinuous-typeengineering problems.Itsperformanceisevaluated bymeansofasetofclassicaltestproblems, and then it is successfully applied to the training of recurrent neural networks to identifylinearandnonlineardynamicplants. AlbertoV.Donati,VinceDarley,andBalaRamachandrandescribetheintegra- tionofanant-basedalgorithmwithagreedyalgorithmforoptimizingtheschedul- ingofamulti-stageplantintheconsumerpackagedgoodsindustry.Thescheduling must provideboth optimal and flexible solutions torespondtofluctuations in de- mand.“Phasetransitions”canbeidentifiedinamultidimensionalspace,whereitis possibletovarythenumberofresourcesavailable. Thefollowing chapter bySung-Soo Kim, Alice E. Smith, and Soon-Jung Hong presentsanantcolonyapproachtooptimallyloadbalancecodedivisionmultipleac- cessmicro-cellular mobilecommunication systems. Loadbalancing isachieved by assigning eachmicro-celltoasector.Thecostfunctionconsidershandoffcostand blockedcallscost,whilethesectorization mustmeetaminimumlevelofcompact- ness.Theproblemisformulatedasaroutingproblemwheretherouteofasingleant createsasectorofmicro-cells.Thereisanantforeachsectorinthesystem,multi- ple ants comprise a colony and multiple colonies operate to find the sectorization withthelowestcost.Themethodisshowntobeeffectiveandhighlyreliable,andis computationallypractical,evenforlargeproblems. GuszEibenandMartijnSchutdiscussnewwaysofcalibratingevolutionaryalgo- rithms,throughasuitablecontroloftheirparameterson-the-fly.Theyfirstreviewthe mainoptionsavailableintheliteratureandpresentsomestatisticsonthemostpop- ularones.Theythenprovidethreecasestudiesindicating thehighpotentialofun- commonvariants.Inparticular,theyrecommendfocusingonparametersregulating selectionandpopulationsize,ratherthanthoseconcerningcrossoverandmutation. ThechapterbyMarcSchoenauer,PierreSavéant, andVincentVidaldescribesa newsequentialhybridizationstrategy,called“Divide-and-Evolve”,thatevolutionarily buildsasequentialslicingoftheproblemathandintoseveral,hopefullyeasier,sub- Preface VII problems. Theembedded (meta-)heuristic is only asked tosolve the “small” prob- lems.Divide-and-Evolve isthusabletogloballysolveproblemsthatareintractable whenfeddirectlyintotheheuristic.Aprominentadvantageofthisapproachisthat it opens up an avenue for multi-objective optimization, even whenusing asingle- objectiveembeddedalgorithm. In their chapter, Carlos García-Martínez and Manuel Lozano are interested in local search based ona genetic algorithm (GA).Theyproposeabinary-coded GA thatappliesacrowdingreplacementmethodinordertokeep,withinthepopulation, differentnicheswithhigh-qualitysolutions.Localsearchcanthenbeperformedby orientatingthesearchinthenearestnichestoasolutionofinterest.ThelocalGAde- signedconsistentlyoutperformedseverallocalsearchproceduresfromtheliterature. ThechapterbyFranciscoB.Pereira,JorgeM.C.Marques,TiagoLeitão,andJorge Tavarespresentsastudyonlocalityinhybridevolutionaryclusteroptimization.Since aclusterisdefinedasanaggregateofbetweenafewandmanymillionsofatomsor molecules,theproblemistofindthearrangementoftheparticlesthatcorresponds tothelowestenergy.Theauthorsarguethatlocalityisanimportantrequisiteofevo- lutionarycomputationtoensuretheefficientsearchforagloballyoptimalsolution. In the following chapter, Pankaj Kumar, Ankur Gupta, Rajshekhar, Valadi K. Jayaraman, and Bhaskar D. Kulkarni present a genetic algorithm-based learning methodologyforclassificationofbenchmarktimeseriesproblemsinmedicaldiag- nosisandprocessfaultdetection. Theresultsindicatethattheconstrained window warpingmethodwithgeneticallylearnedmultiplebandscouldbereliablyemployed foravarietyofclassificationandclusteringproblems. ThechapterbyJong-HwanKim,Chi-HoLee,Kang-HeeLee,andNaveenS.Kup- puswamy focuses on evolving the personality of an artificial creature by using its computer-codedgenomesandevolutionaryGAinasimulatedenvironment.Thear- tificial creature,namedRity,isdevelopedinaDvirtualworldtoobservetheout- comeofitsreactions,accordingtoitsgenome(personality)obtainedthroughevolu- tion. In their chapter, Antonin Ponsich, Catherine Azzaro-Pantel, Serge Domenech, andLucPibouleauprovidesomeguidelinesforGAimplementationinmixedinteger nonlinear programming problems. The support for the work is the optimal batch plantdesign.ThisstudydealswiththetwomainissuesforaGA,i.e.,theprocessing ofcontinuousvariablesbyspecificencodingandthehandlingofconstraints. MárciaMarcondes Altimari Samed and MauroAntonio daSilva SaRavagnani presentapproachesbasedonGAstosolvetheeconomicdispatchproblem.Toelim- inate the cost of the preliminary tuning of parameters, they have performed a co- evolutionary hybrid GA whose parameters are adjusted in the course of the opti- mization. EfrénMezura-Montes,EdgarA.Portilla-Flores, CarlosA.CoelloCoello,Jaime Alvarez-Gallegos,andCarlosA.Cruz-Villarthendescribeanevolutionaryapproach tosolvinganovelmechatronicmultiobjectiveoptimizationproblem,namelythatof pinion-rackcontinuouslyvariabletransmission.Boththemechanicalstructureand thecontrollerperformanceareconcurrentlyoptimizedinordertoproducemechan- ical,electronic,andcontrolflexibilityforthedesignedsystem. VIII Preface Inthelastchapterdevotedtoevolutionaryalgorithms,HélcioVieiraJuniorim- plementsaGAforanaeronauticmilitaryapplication.Thegoalistodetermineanop- timalsequenceofflarelaunchsuchthatthesurvivalprobabilityofanaircraftagainst amissileismaximized. Thefourremainingchaptersdealwithmiscellaneousmetaheuristics.First,Jörn Grahl,StefanMinner,andPeterA.N.Bosmanpresentacondensedoverviewofthe theoryandapplication oftheestimationofdistributionalgorithms (EDAs)inboth the discrete and the continuous problem domain. What differentiates EDAs from otherevolutionaryandnon-evolutionaryoptimizersisthattheyreplacefixedvaria- tionoperatorslikecrossoverandmutationwithadaptivevariationthatcomesfrom aprobabilistic model.EDAshavebeensuccessfully applied tomanyproblemsthat arenotoriouslyhardforstandardgeneticalgorithms. Inthefollowingchapter,ZhenyuYang,JingsongHe,andXinYaoproposeNeigh- borhoodSearch(NS)tobeembeddedwithDifferentialEvolution(DE).Theadvan- tages of NSstrategy in DE are analyzed theoretically. These analyses focus mainly on the change in search step size and population diversity after using neighbor- hood search. Experimental results have shown that DE with neighborhood search hassignificantadvantages overotherexistingalgorithmsinabroadrangeofdiffer- entbenchmarkfunctions.Thescalabilityofthenewalgorithmisalsoevaluatedina numberofbenchmarkproblems,whosedimensionsrangefromto. Sébastien Aupetit, Nicolas Monmarché, and Mohamed Slimane are interested in thetraining ofHiddenMarkovModels(HMMs) using population-based meta- heuristics.Theyhighlighttheuseofthreemethods(GA,antcolonyalgorithm, and particle swarm optimization) with andwithout alocaloptimizer. Thestudyisfirst performedfromatheoretical pointofview;theresultsofexperimentsondifferent setsofartificialandrealdataarethendiscussed. The last chapter of the book by Fred Glover deals with inequalities and target objectives that were recently introduced to guide the search in adaptive memory and evolutionary metaheuristics for mixed integer programming. These guidance approachesareusefulinintensificationanddiversificationstrategiesrelatedtofixing subsetsofvariablesatparticularvalues,andinstrategiesthatuselinearprogramming togeneratetrialsolutionswhosevariablesareinducedtoreceiveintegervalues.The authorshowshowtoimprovesuchapproachesinthecaseof-mixedintegerpro- gramming. Wedohopeyouwillfindthevolumeinterestingandthoughtprovoking.Enjoy! Adelaide,Australia ZbigniewMichalewicz Paris,France PatrickSiarry March ListofContents Comparison of Simulated Annealing, Interval Partitioning andHybridAlgorithmsinConstrainedGlobalOptimization C.S.Pedamallu,L.Özdamar............................................  Four-bar Mechanism Synthesis for n Desired Path Points UsingSimulatedAnnealing H.Martínez-Alfaro ...................................................  “MOSS-II”Tabu/ScatterSearchforNonlinearMultiobjectiveOptimization R.P.Beausoleil .......................................................  FeatureSelectionforHeterogeneousEnsembles ofNearest-neighbourClassifiersUsingHybridTabuSearch M.A.Tahir,J.E.Smith .................................................  A Parallel Ant Colony Optimization Algorithm Based onCrossoverOperation A.Kalinli,F.Sarikoc ..................................................  AnAnt-biddingAlgorithmforMultistageFlowshopSchedulingProblem: OptimizationandPhaseTransitions A.V.Donati,V.Darley,B.Ramachandran.................................  Dynamic Load Balancing Using an Ant Colony Approach inMicro-cellularMobileCommunicationsSystems S.-S.Kim,A.E.Smith,S.-J.Hong ........................................  NewWaystoCalibrateEvolutionaryAlgorithms A.E.Eiben,M.C.Schut ................................................  Divide-and-Evolve: a Sequential Hybridization Strategy UsingEvolutionaryAlgorithms M.Schoenauer,P.Savéant,V.Vidal......................................  X ListofContents LocalSearchBasedonGeneticAlgorithms C.García-MartínezandM.Lozano......................................  Designing EfficientEvolutionary Algorithms forClusterOptimization: AStudyonLocality F.B.Pereira,J.M.C.Marques,T.Leitão,J.Tavares ..........................  Aligning TimeSerieswithGenetically TunedDynamicTimeWarping Algorithm P.Kumar,A.Gupta,Rajshekhar,V.K.Jayaraman,B.D.Kulkarni ..............  Evolutionary Generation of Artificial Creature’s Personality forUbiquitousServices J-H.Kim,C-H.Lee,K-H.Lee,N.S.Kuppuswamy ..........................  Some Guidelines for Genetic Algorithm Implementation inMINLPBatchPlantDesignProblems A.Ponsich,C.Azzaro-Pantel,S.Domenech,L.Pibouleau....................  CoevolutionaryGeneticAlgorithmtoSolveEconomicDispatch M.M.A.Samed,M.A.daS.S.Ravagnani..................................  AnEvolutionary ApproachtoSolveaNovelMechatronicMultiobjective OptimizationProblem E.Mezura-Montes,E.A.Portilla-Flores,C.A.CoelloCoello,J.Alvarez-Gallegos, C.A.Cruz-Villar......................................................  Optimizing Stochastic Functions Using a Genetic Algorithm: AnAeronauticMilitaryApplication H.V.Junior ..........................................................  Learning Structure Illuminates Black Boxes – An Introduction toEstimationofDistributionAlgorithms J.Grahl,S.Minner,P.A.N.Bosman ......................................  MakingaDifferencetoDifferentialEvolution Z.Yang,J.He,X.Yao..................................................  HiddenMarkovModelsTrainingUsingPopulation-basedMetaheuristics S.Aupetit,N.Monmarché,M.Slimane ...................................  Inequalities and Target Objectives for Metaheuristic Search –PartI: MixedBinaryOptimization F.Glover ............................................................  Index ..............................................................  ListofContributors JaimeAlvarez-Gallegos CarlosA.CoelloCoello Cinvestav-Ipn Cinvestav-Ipn DepartamentodeIngenieriaEléctrica DepartamentodeComputación MéxicoD.F.,Mexico MéxicoD.F.,Mexico [email protected] [email protected] SébastienAupetit CarlosA.Cruz-Villar UniversitédeTours Cinvestav-Ipn Laboratoired’Informatique DepartamentodeIngenieriaEléctrica Tours,France MéxicoD.F.,Mexico [email protected] [email protected] CatherineAzzaro-Pantel VinceDarley Laboratoire deGénieChimiquede EurobiosUKLtd. Toulouse LondonECAAB,UK Toulouse,France [email protected] [email protected] SergeDomenech RicardoP.Beausoleil Laboratoire deGénieChimiquede InstitutodeCiberneticaMatematicay Toulouse Fisica Toulouse,France DepartmentodeOptimizacion CiudadHabana,Cuba [email protected] [email protected] AlbertoV.Donati PeterA.N.Bosman JointResearchCenter CentrumvoorWiskundeenInformatica EuropeanCommission GBAmsterdam,TheNetherlands Ispra(VA),Italy [email protected] [email protected]

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.