Wolfgang Stolzmann - Böcker

This book constitutes the refereed proceedings of the 5th International Workshop on Learning Classifier Systems, IWLCS 2003, held in Granada, Spain in September 2003 in conjunction with PPSN VII.The 10 revised full papers presented together with a comprehensive bibliography on learning classifier systems were carefully reviewed and selected during two rounds of refereeing and improvement. All relevant issues in the area are addressed.

This book constitutes the thoroughly refereed post-proceedings of the Third International Workshop on Learning Classifier Systems, IWLCS 2000, held in Paris, France in September 2000. The 13 revised full papers presented have gone through two rounds of reviewing and selection. Also included is a comprehensive LCS bibliography listing more than 600 entries as well as an appendix. The papers are organized in topical sections on theory, applications, and advanced architectures.

The Fourth International Workshop on Learning Classi?er Systems (IWLCS 2001)washeldJuly7-8,2001,inSanFrancisco,California,duringtheGenetic andEvolutionaryComputationConference(GECCO2001). Wehaveincluded inthisvolumerevisedandextendedversionsofelevenofthepaperspresented attheworkshop. Thevolumeisorganizedintotwomainparts. The?rstisdedicatedtoimportant theoreticalissuesoflearningclassi?ersystemsresearchincludingthein?uence ofexplorationstrategy,amodelofself-adaptiveclassi?ersystems,andtheuse ofclassi?ersystemsforsocialsimulation. Thesecondpartcontainspapersd- cussing applications of learning classi?er systems such as data mining, stock trading,andpowerdistributionnetworks. AnappendixcontainsapaperpresentingaformaldescriptionofACS,arapidly emerginglearningclassi?ersystemmodel. Thisbookistheidealcontinuationofthetwovolumesfromthepreviouswo- shops,publishedbySpringer-VerlagasLNAI1813andLNAI1996. Wehopeit willbeausefulsupportforresearchersinterestedinlearningclassi?ersystems andwillprovideinsightsintothemostrelevanttopicsandthemostinteresting openissues. April2002 PierLucaLanzi WolfgangStolzmann StewartW. Wilson Organization The Fourth International Workshop on Learning Classi?er Systems (IWLCS 2001)washeldJuly7-8,2001inSanFrancisco(CA),USA,duringtheGenetic andEvolutionaryConference(GECCO2001).OrganizingCommittee PierLucaLanzi PolitecnicodiMilano,Italy WolfgangStolzmann DaimlerChryslerAG,Germany StewartW. Wilson TheUniversityofIllinoisatUrbana-Champaign,USA PredictionDynamics,USA ProgramCommittee ErikBaum NECResearchInstitute,USA AndreaBonarini PolitecnicodiMilano,Italy LashonB. Booker TheMITRECorporation,USA MartinV. Butz UniversityofWur .. zburg,Germany LawrenceDavis NuTechSolutions,USA TerryFogarty SouthbankUniversity,UK JohnH. Holmes UniversityofPennsylvania,USA TimKovacs UniversityofBirmingham,UK PierLucaLanzi PolitecnicodiMilano,Italy RickL. Riolo UniversityofMichigan,USA OlivierSigaud AnimatLab-LIP6,France RobertE. Smith TheUniversityofTheWestofEngland,UK WolfgangStolzmann DaimlerChryslerAG,Germany KeikiTakadama ATRInternational,Japan StewartW. Wilson TheUniversityofIllinoisatUrbana-Champaign,USA PredictionDynamics,USA TableofContents ITheory BiasingExplorationinanAnticipatoryLearningClassi?erSystem ...3 MartinV. Butz An Incremental Multiplexer Problem and Its Uses in Classi?er System Research...23 LawrenceDavis,ChunshengFu,StewartW. Wilson AMinimalModelofCommunicationforaMulti-agentClassi?erSystem. .32 ' GillesEn'ee,CathyEscazut A Representation for Accuracy-Based Assessment of Classi?er System PredictionPerformance...43 JohnH. Holmes ASelf-AdaptiveXCS...57 JacobHurst,LarryBull TwoViewsofClassi?erSystems ...74 TimKovacs SocialSimulationUsingaMulti-agentModelBasedonClassi?erSystems: TheEmergenceofVacillatingBehaviourinthe"ElFarol"BarProblem...88 LuisMiramontesHercog,TerenceC. Fogarty II Applications XCSandGALE:AComparativeStudyofTwoLearningClassi?erSystems onDataMining...115 EsterBernad'o,XavierLlor'a,JosepM. Garrell APreliminaryInvestigationofModi?edXCSasaGenericDataMining Tool...133 PhillipWilliamDixon,DavidW. Corne,MartinJohnOates ExplorationsinLCSModelsofStockTrading ...151 SoniaSchulenburg,PeterRoss On-LineApproachforLossReductioninElectricPowerDistribution NetworksUsingLearningClassi?erSystems...181 Patr'?ciaAmancioVargas,ChristianoLyraFilho, FernandoJ. VonZuben VIII TableofContents CompactRulesetsfromXCSI ...197 StewartW. Wilson III Appendix AnAlgorithmicDescriptionofACS2 ...211 MartinV. Butz,WolfgangStolzmann AuthorIndex ...231 BiasingExplorationinan AnticipatoryLearningClassi?erSystem MartinV. Butz DepartmentofCognitivePsychology,UniversityofWurz .. burg R.. ontgenring11,97070Wurz .. burg,Germany butz@psychologie. uni-wuerzburg. de Abstract. Thechapterinvestigateshowmodelandbehaviorallearning can be improved in an anticipatory learning classi?er system by bi- ing exploration. First, theappliedsystemACS2isexplained. Next,an overviewoverthepossibilitiesofapplyingexplorationbiasesinanant- ipatory learning classi?er systemand speci?cally ACS2 is provided.

Learning Classifier Systems (LCS) are a machine learning paradigm introduced by John Holland in 1976. They are rule-based systems in which learning is viewed as a process of ongoing adaptation to a partially unknown environment through genetic algorithms and temporal difference learning. This book provides a unique survey of the current state of the art of LCS and highlights some of the most promising research directions. The first part presents various views of leading people on what learning classifier systems are. The second part is devoted to advanced topics of current interest, including alternative representations, methods for evaluating rule utility, and extensions to existing classifier system models. The final part is dedicated to promising applications in areas like data mining, medical data analysis, economic trading agents, aircraft maneuvering, and autonomous robotics. An appendix comprising 467 entries provides a comprehensive LCS bibliography.

The work embodied in this volume was presented across three consecutive e- tions of the International Workshop on Learning Classi?er Systems that took place in Chicago (2003), Seattle (2004), and Washington (2005). The Genetic and Evolutionary Computation Conference, the main ACM SIGEvo conference, hosted these three editions. The topics presented in this volume summarize the wide spectrum of interests of the Learning Classi?er Systems (LCS) community. The topics range from theoretical analysis of mechanisms to practical cons- eration for successful application of such techniques to everyday data-mining tasks. When we started editing this volume, we faced the choice of organizing the contents in a purely chronologicalfashion or as a sequence of related topics that help walk the reader across the di?erent areas. In the end we decided to or- nize the contents by area, breaking the time-line a little. This is not a simple endeavor as we can organize the material using multiple criteria. The tax- omy below is our humble e?ort to provide a coherent grouping. Needless to say, some works may fall in more than one category. The four areas are as follows: Knowledge representation. These chapters elaborate on the knowledge r- resentations used in LCS. Knowledge representation is a key issue in any learning system and has implications for what it is possible to learn and what mechanisms shouldbe used. Four chapters analyze di?erent knowledge representations and the LCS methods used to manipulate them.