Knowledge Acquisition: Approaches, Algorithms and Applications: Pacific Rim Knowledge Acquisition Workshop, PKAW 2008, Hanoi, Vietnam, December 15-16, 2008, Revised Selected Papers PDF
Preview Knowledge Acquisition: Approaches, Algorithms and Applications: Pacific Rim Knowledge Acquisition Workshop, PKAW 2008, Hanoi, Vietnam, December 15-16, 2008, Revised Selected Papers
Lecture Notes in Artificial Intelligence 5465 EditedbyR.Goebel,J.Siekmann,andW.Wahlster Subseries of Lecture Notes in Computer Science Debbie Richards Byeong-Ho Kang (Eds.) Knowledge Acquisition: Approaches, Algorithms and Applications Pacific Rim KnowledgeAcquisitionWorkshop, PKAW 2008 Hanoi,Vietnam, December 15-16, 2008 Revised Selected Papers 1 3 SeriesEditors RandyGoebel,UniversityofAlberta,Edmonton,Canada JörgSiekmann,UniversityofSaarland,Saarbrücken,Germany WolfgangWahlster,DFKIandUniversityofSaarland,Saarbrücken,Germany VolumeEditors DebbieRichards MacquarieUniversity,ComputingDepartment DivisionofInformationandCommunicationSciences Sydney,NSW,2109,Australia E-mail:[email protected] Byeong-HoKang UniversityofTasmania SchoolofComputingandInformationSystems Launceston,TAS7250,Australia E-mail:[email protected] LibraryofCongressControlNumber:Appliedfor CRSubjectClassification(1998):I.2.6,I.2,H.2.8,H.3-5,F.2.2,C.2.4,K.3 LNCSSublibrary:SL7–ArtificialIntelligence ISSN 0302-9743 ISBN-10 3-642-01714-2SpringerBerlinHeidelbergNewYork ISBN-13 978-3-642-01714-8SpringerBerlinHeidelbergNewYork Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialis concerned,specificallytherightsoftranslation,reprinting,re-useofillustrations,recitation,broadcasting, reproductiononmicrofilmsorinanyotherway,andstorageindatabanks.Duplicationofthispublication orpartsthereofispermittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9,1965, initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer.Violationsareliable toprosecutionundertheGermanCopyrightLaw. springer.com ©Springer-VerlagBerlinHeidelberg2009 PrintedinGermany Typesetting:Camera-readybyauthor,dataconversionbyScientificPublishingServices,Chennai,India Printedonacid-freepaper SPIN:12676811 06/3180 543210 Preface Withthegrowingrecognitionofthepivotalrolethatknowledgeplaysinthesus- tainability,competitivenessandgrowthofindividuals,organizationsandsociety, finding solutions to address the knowledge acquisition bottleneck is even more important today than in the early stages of this field. The knowledge acquisi- tion community is interested in topics spanning from the fundamental views on knowledgethataffecttheknowledgeacquisitionprocessandtheuseofknowledge in knowledge engineering to the evaluation of knowledge acquisitiontechniques, tools and methods. Asafieldwithinthelargerfieldofartificialintelligence(AI),solutionsincor- porating other areas of AI such as ontologicalengineering, agent-basedtechnol- ogy, robotics, image recognition and the Semantic Web are common, as are knowledge acquisition methods related to other fields of computing such as databases,the Internet,informationretrieval,languagetechnology,softwareen- gineering, decision support systems and game technology. Many solutions are application focused addressing real-world problems such as knowledge mainte- nance and validation, reuse and sharing, merging and reconciliation within a wide range of problem domains. The Pacific Knowledge Acquisition Workshops (PKAW) have provided a fo- rum for the past two decades for researchers and practitioners in the Pacific region and beyond who work in the field of knowledge acquisition. PKAW cov- ers a spectrum of techniques and approaches ranging from manual knowledge acquisitionfromahumanexperttofullyautomatedknowledgeacquisitionusing machine-learning or data-mining methods. ThisvolumeseekstodisseminatethelatestsolutionsfromthePacificKnowl- edgeAcquisitionWorkshop2008(PKAW2008)heldinHanoi,TaiwanduringDe- cember15-16,2008inconjunctionwiththePacificRimInternationalConference onArtificialIntelligence (PRICAI 2008).The workshopreceived57submissions from14countries.Fromthese,weaccepted15papers(26%)forfullpresentation andanother14forshortpresentation.Allpaperswereblindreviewedbyatleast three members of the Program Committee. This volume contains a selection of these papersfurther revisedfollowingworkshopdiscussions.The papersdemon- strate a balance of theoretical, technical and application-driven research, many papersincorporatingallthreefoci.Approximatelyhalfthepapersreflectthe in- creasinguseofWeb-baseddataforknowledgediscoveryandmanagement. The Workshop Co-chairs would like to thank all those who were involved in PKAW 2008 including the PRICAI 2008 Organizing Committee, PKAW Pro- gramCommitteemembers,thosewhosubmittedpapersandreviewedthemand VI Preface ofcoursetheauthors,presentersandattendees.Wewarmlyinviteyoutopartic- ipate in PKAW2010anticipated to be held in Seoul,Koreain conjunction with PRICAI 2010. March 2009 Debbie Richards Byeong Ho Kang Organization Honorary Chairs Paul Compton University of New South Wales, Australia Hiroshi Motoda OsakaUniversityandAFOSR/AOARD,Japan Program Co-Chairs Debbie Richards Macquarie University, Australia Byeong Ho Kang University of Tasmania, Australia Program Committee Ghassan Beydoun University of Wollongong, Australia Bob Colomb University of Technology Malaysia, Malaysia Paul Compton University of New South Wales, Australia Richard Dazeley University of Ballarat, Australia Fabrice Guillet LINA - Polytech’Nantes, France Udo Hahn Jena University, Germany Byeong Ho Kang University of Tasmania, Australia Rob Kremer University of Calgary, Canada Maria Lee Shih Chien University, Taiwan Tim Menzies LCSEE, WVU, USA Toshiro Minami Kyushu University, Japan Hiroshi Motoda OsakaUniversityandAFOSR/AOARD,Japan Kozo Ohara Osaka University, Japan Takashi Okada Kwansei Gakuin University, Japan Son Pham College of Technology, VNU, Vietnam Frank Puppe University of Wu¨rzburg, Germany Ulrich Reimer University of Applied Sciences St. Gallen, Switzerland Debbie Richards Macquarie University, Australia Hendra Suryanto University of New South Wales, Australia Seiji Yamada National Institute of Informatics, Japan Kenichi Yoshida University of Tsukuba, Japan Tetsuya Yoshida Hokkaido Univerisity, Japan VIII Organization Additional Reviewers Daniel Bidulock Jason Heard Uwe Heck Claudia Marinica Christian Thiel Ryan Yee Table of Contents Machine Learning and Data Mining Experiments with Adaptive Transfer Rate in Reinforcement Learning... 1 Yann Chevaleyre, Aydano Machado Pamponet, and Jean-Daniel Zucker Clustering over Evolving Data Streams Based on Online Recent-Biased Approximation .................................................. 12 Wei Fan, Yusuke Koyanagi, Koichi Asakura, and Toyohide Watanabe Automatic Database Creation and Object’s Model Learning ........... 27 Nguyen Dang Binh and Thuy Thi Nguyen Finding the Most Interesting Association Rules by Aggregating Objective Interestingness Measures................................. 40 Tri Thanh Nguyen Le, Hiep Xuan Huynh, and Fabrice Guillet Pruning Strategies Based on the Upper Bound of Information Gain for Discriminative Subgraph Mining ................................... 50 Kouzou Ohara, Masahiro Hara, Kiyoto Takabayashi, Hiroshi Motoda, and Takashi Washio A Novel Classification Algorithm Based on Association Rules Mining ... 61 Bay Vo and Bac Le Incremental Knowledge Acquisition Multiple Classification Ripple Round Rules: A Preliminary Study ...... 76 Ivan Bindoff, Tristan Ling, and Byeong-Ho Kang Generalising Symbolic Knowledge in Online Classification and Prediction ...................................................... 91 Richard Dazeley and Byeong-Ho Kang Web-Based Techniques and Application Using Formal Concept Analysis towards Cooperative E-Learning....... 109 Ghassan Beydoun Navigation and Annotation with Formal Concept Analysis (Extended Abstract) ....................................................... 118 Peter Eklund and Jon Ducrou X Table of Contents What Does an Information Diffusion Model Tell about Social Network Structure? ...................................................... 122 Takayasu Fushimi, Takashi Kawazoe, Kazumi Saito, Masahiro Kimura, and Hiroshi Motoda Web Mining for Malaysia’s Political Social Networks Using Artificial Immune System ................................................. 137 Ahmad Nadzri Muhammad Nasir, Ali Selamat, and Md. Hafiz Selamat Accessing Touristic Knowledge Bases through a Natural Language Interface........................................................ 147 Juana Mar´ıa Ruiz-Mart´ınez, Dagoberto Castellanos-Nieves, Rafael Valencia-Garc´ıa, Jesualdo Toma´s Ferna´ndez-Breis, Francisco Garc´ıa-Sa´nchez, Pedro Jos´e Vivancos-Vicente, Juan Salvador Castej´on-Garrido, Juan Bosco Camo´n, and Rodrigo Mart´ınez-B´ejar ItemSpider: Social Networking Service That Extracts Personal Character from Individual’s Book Information ....................... 161 Tetsuya Tsukamoto, Hiroyuki Nishiyama, and Hayato Ohwada Acquiring Marketing Knowledge from Internet Bulletin Boards ........ 173 Hiroshi Uehara and Kenichi Yoshida Domain Specific Knowledge Acquisition Methods and Applications A Design for Library Marketing System and Its Possible Applications..................................................... 183 Toshiro Minami Knowledge Audit on Special Children Communities .................. 198 Aida Suzana Sukiam, Azizah Abdul Rahman, and Wardah Zainal Abidin A Dance Synthesis System Using Motion Capture Data............... 208 Kenichi Takahashi and Hiroaki Ueda Discovering Areas of Expertise from Publication Data ................ 218 Meredith Taylor and Debbie Richards Facial Feature Extraction Using Geometric Feature and Independent Component Analysis ............................................. 231 Toan Thanh Do and Thai Hoang Le Author Index.................................................. 243 Experiments with Adaptive Transfer Rate in Reinforcement Learning Yann Chevaleyre1, Aydano Machado Pamponet2, and Jean-Daniel Zucker3 1 Universit Paris-Dauphine [email protected] 2 Universit Paris 6 [email protected] 3 IRD URGodes Centre IRD del’Ile deFrance, Bondy,France [email protected] Abstract. Transfer algorithms allow the use of knowledge previously learned on related tasks to speed-up learning of the current task. Re- cently, many complex reinforcement learning problems have been suc- cessfully solved by efficient transfer learners. However, most of these algorithmssufferfromasevereflaw:theyareimplicitlytunedtotransfer knowledge between tasks having a given degree of similarity. In other words, if the previous task is very dissimilar (resp. nearly identical) to the current task, then the transfer process might slow down the learn- ing(resp.mightbefarfromoptimalspeed-up).Inthispaper,weaddress thisspecificissuebyexplicitlyoptimizingthetransferratebetweentasks and answer to the question : “can the transfer rate be accurately opti- mized, and at what cost ?”. We show that this optimization problem is related to the continuum bandit problem. We then propose a generic adaptivetransfermethod(AdaTran),whichallows toextendseveralex- istingtransfer learningalgorithms tooptimize thetransferrate.Finally, we run several experimentsvalidating our approach. 1 Introduction In the reinforcement learning problem, an agent acts in an unknown environ- ment, with the goal of maximizing its reward. All learning agents have to face the exploration-exploitation dilemma: whether to act so as to explore unknown areas or to act consistently with experience to maximize reward(exploit). Most research on reinforcement learning deal with this issue. Recently Strehl et al. [17]showedthatnear optimalstrategiescouldbe reachedin as few asO(cid:2)(S×A) time steps. However, in many real-world learning problems, the state space or the action space have an exponential size. One way to circumvent this problem is to use previously acquired knowledge related to the current task being learned. This knowledge may then be used to guideexplorationthroughthestate-actionspace,hopefullyleadingtheagentto- wardsareasinwhichhighrewardscanbefound.Thisknowledgecanbeacquired in different ways: D.RichardsandB.-H.Kang(Eds.):PKAW2008,LNAI5465,pp.1–11,2009. (cid:2)c Springer-VerlagBerlinHeidelberg2009
Description:The list of books you might like
