Fuzzy cognitive maps for applied sciences and engineering : from fundamentals to extensions and learning algorithms PDF
Preview Fuzzy cognitive maps for applied sciences and engineering : from fundamentals to extensions and learning algorithms
Intelligent Systems Reference Library 54 Elpiniki I. Papageorgiou Editor Fuzzy Cognitive Maps for Applied Sciences and Engineering From Fundamentals to Extensions and Learning Algorithms Intelligent Systems Reference Library Volume 54 Series Editors J. Kacprzyk, Warsaw, Poland L. C. Jain, Canberra, Australia For furthervolumes: http://www.springer.com/series/8578 Elpiniki I. Papageorgiou Editor Fuzzy Cognitive Maps for Applied Sciences and Engineering From Fundamentals to Extensions and Learning Algorithms 123 Editor ElpinikiI.Papageorgiou Department of Computer Engineering Technological Educational Instituteof CentralGreece Lamia Greece Additionalmaterialtothisbookcanbedownloadedfromhttp://extras.springer.com. ISSN 1868-4394 ISSN 1868-4408 (electronic) ISBN 978-3-642-39738-7 ISBN 978-3-642-39739-4 (eBook) DOI 10.1007/978-3-642-39739-4 SpringerHeidelbergNewYorkDordrechtLondon LibraryofCongressControlNumber:2013950727 (cid:2)Springer-VerlagBerlinHeidelberg2014 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionor informationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purposeofbeingenteredandexecutedonacomputersystem,forexclusiveusebythepurchaserofthe work. Duplication of this publication or parts thereof is permitted only under the provisions of theCopyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the CopyrightClearanceCenter.ViolationsareliabletoprosecutionundertherespectiveCopyrightLaw. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexempt fromtherelevantprotectivelawsandregulationsandthereforefreeforgeneraluse. While the advice and information in this book are believed to be true and accurate at the date of publication,neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityfor anyerrorsoromissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,with respecttothematerialcontainedherein. Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) To my husband Nikos for his patience all these years and To Yiannis and Alexandros, the two suns of my life Foreword Prof. Elipiniki I. Papageorgiou has organized and edited an important new con- tribution to the rapidly growing field offuzzy cognitive maps (FCMs). This new volume further extends the practical and theoretical boundaries of this interna- tional FCM research effort. It includes several predictive FCM models that range from rulebased systems to adaptive dynamical systems with their own learning laws. ThesediversemodelssuggestthatFCMswillplayimportantrolesinthefuture of both computing and machine intelligence. This includes applications to so- called ‘‘big data’’ and what we can here call ‘‘big knowledge.’’ FCMsarenaturaltoolstoprocessbigdata.Theirgraphicalstructureofacyclic signed directed graph allows the user to specify coarse or fine levels of causal granularitythroughthechoice ofconceptnodesandcausaledges.Controllingthe causalgranularitycancombatthesystemicproblemofexponentialruleexplosion or the curse of dimensionality that infests large rule-based systems and semantic networks. Controlled FCM granularity results in a coarse or fine rule-based compression of the streaming data because the directed causal links define fuzzy if–then rules. Granularized FCMs can use statistical learning laws that scale with the data stream itself. Such adaptive FCMs can gradually update the causal links and can form or split or delete concept nodes as the data streams into the system (although the FCM field still needs a good data-based theory of concept-node formation). This learning process can go on indefinitely. Users can also add or delete FCM elements at any time in the learning process. A user or higher level adaptive system can adjust the learning-rate parameters to match the flow of the data. The resulting FCM at any given time always gives high-level causal and predictive insight into the data flux. FCMs alsooffer anatural representational frameworkfor what we can callbig knowledge—the world’s vast and growing body of expert analysis and advice. Thisstructuredknowledgeoftentakesthetraditionalformofbooksortechnical articles or essays. But it can also take the form of Internet blogs or media inter- views or expert-witness court transcripts. Big knowledge includes the whole vii viii Foreword panoply offixed or oral knowledge that the law calls documentary or testimonial evidence. Today that knowledge exists largely in disconnected sources or chunks aroundtheworld.FCMshavetheabilitytoconnectandcombinetheseknowledge chunks into a unified framework for policy and engineering analysis and for fur- ther computer and knowledge processing. FCMs can synthesize this disparate knowledge through a simple transform technique involving connection or adjacency matrices. The technique resembles how a Fourier transform converts a time signal into the frequency domain where the user can filter or modify the signal and then inverse-transform the result back to the time domain. An FCM can represent each knowledge chunk or expert contribution. Then we can translate each such FCM knowledge chunk into an augmentedsquareconnectionmatrixconformableforaddition.Thatinturnallows the formation of a massive knowledge base by appropriately weighting and combiningthematricesintoalargesparsematrix.Thentranslatingthematrixback into an FCM causal digraph gives the final knowledge base as a massive FCM. This fusion of all structured knowledge amounts to a worldwide FCM knowledge-representation project. This long-term effort will be the direct benefi- ciary of Google Books and the Gutenberg Project and Wikisource and related large-scale efforts to digitize and make available the world’s text-based docu- mentary evidence. Every book chapter or essay should have its own FCM instantiation.SofartherehavebeenafewmanualeffortsatsuchFCMknowledge translation and synthesis. That includes some of the FCMs developed in this volume. But a fully automated FCM synthesizer remains a research goal for the future. FCMs can advance big knowledge in yet another way: they can naturally represent deep knowledge in stacked or multilayered FCMs. These multilayer structures are far more complex and expressive than stacked or deep neural net- works. Almost all such multilayered neural networks have only a feedforward architectureandthustheyhaveonlytrivialdynamics.Theyhavenoconnectionsat allamongtheneuronsinagivenvisiblelayerorhiddenlayer.Nordothesynaptic edges or most neurons have any meaningful interpretation. So these minimal multilayer structures allow little more than blind statistical training of the con- tiguous layers and of the overall network itself. But an FCM’s representation power and rich feedback dynamics stem directly from the cyclic causal edge connections among the concept nodes in a layer—cycles that undermine most traditional expert systems and Bayesian networks. Stacked FCMs can represent knowledgeondifferenttimescalesbothwithinFCMlayersandespeciallybetween FCMlayers.ThesestackedFCMsalsodonotneedtofunctiononlyinfeedforward mode.Theycanallowfeedbackfromhigherlayerstolowerlayersanddosoagain ondifferenttimescales.ThustheentirestackedormultilayerFCMcanreverberate as it passes through successive dynamical equilibria. Concept nodes in a given Foreword ix FCMlayercanalsobranchlaterallytootherFCMsoreventootherstackedFCMs. Hence they too can fuse or combine with other FCMs to produce ever larger connected knowledge bases. Thesearenear-termandlong-termgoalsforFCMresearch.Thepresentvolume does an excellent job of moving in those and other directions as well as demon- strating the analytic and predictive power of FCM-based knowledge engineering. Bart Kosko Professor of Electrical Engineering and Law University of Southern California Los Angeles USA Contents 1 Methods and Algorithms for Fuzzy Cognitive Map-based Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Elpiniki I. Papageorgiou and Jose L. Salmeron 2 Fuzzy Cognitive Maps as Representations of Mental Models and Group Beliefs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 S. A. Gray, E. Zanre and S. R. J. Gray 3 FCM Relationship Modeling for Engineering Systems. . . . . . . . . . 49 O. Motlagh, S. H. Tang, F. A. Jafar and W. Khaksar 4 Using RuleML for Representing and Prolog for Simulating Fuzzy Cognitive Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Athanasios Tsadiras and Nick Bassiliades 5 Fuzzy Web Knowledge Aggregation, Representation, and Reasoning for Online Privacy and Reputation Management . . . . . 89 Edy Portmann and Witold Pedrycz 6 Decision Making by Rule-Based Fuzzy Cognitive Maps: An Approach to Implement Student-Centered Education. . . . . . . . 107 A. Peña-Ayala and J. H. Sossa-Azuela 7 Extended Evolutionary Learning of Fuzzy Cognitive Maps for the Prediction of Multivariate Time-Series . . . . . . . . . . . . . . . 121 Wojciech Froelich and Elpiniki I. Papageorgiou 8 Synthesis and Analysis of Multi-Step Learning Algorithms for Fuzzy Cognitive Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Alexander Yastrebov and Katarzyna Piotrowska xi xii Contents 9 Designing and Training Relational Fuzzy Cognitive Maps . . . . . . 145 Grzegorz Słon´ and Alexander Yastrebov 10 Cooperative Autonomous Agents Based on Dynamical Fuzzy Cognitive Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Márcio Mendonça, Lúcia Valéria Ramos de Arruda and Flávio Neves-Jr 11 FCM-GUI: A Graphical User Interface for Big Bang-Big Crunch Learning of FCM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 Engin Yesil, Leon Urbas and Anday Demirsoy 12 JFCM : A Java Library for Fuzzy Cognitive Maps. . . . . . . . . . . 199 Dimitri De Franciscis 13 Use and Evaluation of FCM as a Tool for Long Term Socio Ecological Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221 Martin Wildenberg, Michael Bachhofer, Kirsten G. Q. Isak and Flemming Skov 14 Using Fuzzy Grey Cognitive Maps for Industrial Processes Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 Jose L. Salmeron and Elpiniki I. Papageorgiou 15 Use and Perspectives of Fuzzy Cognitive Maps in Robotics . . . . . 253 Ján Vašcˇák and Napoleon H. Reyes 16 Fuzzy Cognitive Maps for Structural Damage Detection . . . . . . . 267 Ranjan Ganguli 17 Fuzzy Cognitive Strategic Maps . . . . . . . . . . . . . . . . . . . . . . . . . 291 M. Glykas 18 The Complex Nature of Migration at a Conceptual Level: An Overlook of the Internal Migration Experience of Gebze Through Fuzzy Cognitive Mapping Method . . . . . . . . . . . . . . . . 319 Tolga Tezcan 19 Understanding Public Participation and Perceptions of Stakeholders for a Better Management in Danube Delta Biosphere Reserve (Romania). . . . . . . . . . . . . . . . . . . . . . . . . . . 355 M. N. Va˘idianu, M. C. Adamescu, M. Wildenberg and C. Tetelea
The list of books you might like
