Lecture Notes in Artificial Intelligence 1188 Subseries of Lecture Notes in Computer Science Edited by J. G. Carbonell and J. Siekmann Lecture Notes in Computer Science Edited by G. Goos, J. Hartmanis and J. van Leeuwen
Trevor E Martin Anca L. Ralescu (Eds.) Fuzzy Logic in Artificial Intelligence Towards Intelligent Systems IJCAI '95 Workshop Montrdal, Canada, August l 9-21, 1995 Selected Papers Springer
Series Editors Jaime G. Carbonell, Carnegie Mellon Universit2~ Pittsburgh, PA, USA J0rg Siekmann, University of Saarland, Saarbriicken, Germany Volume Editors Trevor P. Martin Advanced Computing Research Centre, University of Bristol Bristol BS8 1TR, United Kingdom E-mail: trevor.martin @bristol.ac.uk Anca L. Ralescu Department of Electrical and Computer Engineering and Computer Science University of Cincinnati Cincinnati, Ohio 45221-0030, USA E-mail: anca.ralescu@uc.edu Cataloging-in-Publication Data applied for Die Deutsche Bibliothek- CIP-Einheitsaufmdlme Fuzzy logic in artiflcal intelligence : towards intelligent systems ; selected papers I IJCAI '95 workshop, Montreal, Canada, August 19-21, 1995. Trevor Martin ;/inca L. Raleseu (od.). -Berlin ; Heidelberg ; New York ; Barcelona ; Budapest ; Hong Kong; London ; Milan ; Paris ; Santa Clara; Singapore ; Tokyo : Springer, I996 (i.,ectm~ aotes in computer science ; "Col. 1188 : Lecture notes in anifiei~ intellis(race) ISBlq 3-540.62474.0 N'B: Martin, Tr~mr [Hrs~l; I/CAI <14, 1996, M(m~; GT CR Subject Classification (1991): 1.2,F.4.1, 1.5.1, H.4.2, J.2 ISBN 3-540-62474-0 Springer-Verlag Berlin Heidelberg New York This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer ~ Violations are liable for prosecution under the German Copyright Law. 9 Springer-Verlag Berlin Heidelberg 1997 Printed in Germany Typesetting: Camera ready by author SPIN 10549933 06/3142-5 4 3 2 1 0 Printed on acid-free paper
Foreword Fuzzy logic based methods have become increasingly popular in recent years. Books, conferences, and a large number of journal articles point to this. But more importantly, all over the world, practical implementations of algorithms using fuzzy logic in a variety of domains has contributed to the realization of a new generation of smart products. It can be safely stated that the first proof of the potential of fuzzy logic came with the implementation of fuzzy logic control. Fuzzy control has seen enormous success in the past decade, with applications as diverse as washing machines, air conditioners, elevator control, subway systems, cruise control, etc. This wide range of commercial applications and a host of less publicized industrial products used by other manufacturers has firmly established the need for fuzzy set theory in modeling the real world. What is its contribution, if any, to the field of intelligent systems? At one end of the spectrum, we have simple control applications -- simple in the sense that they take one or two sensor inputs and set a control value as output, following a one-step reasoning scheme based on fuzzy rules. These rules are if-then statements in which the condition and conclusion parts are expressed using fuzzy predicates, that is, predicates represented as fuzzy sets. Fuzzy predicates have given the possibility to better express heuristic knowledge that a system designer might have about the system as well as the ability to build into the system a tolerance for imprecision in the data on which the system operates. Such applications would normally fall into the domain of control engineering. At the other end of the spectrum, we have complex applications such as health management, image understanding, and foreign exchange dealing. These are normally considered to be in the domain of artificial intelligence; however, AI research has tended to concentrate on symbolic and logical reasoning, often neglecting uncertainty and what we may call the messier aspects of the real world, and focusing instead on theoretical elegance. The comfort of working within a theory iswell known. The difficulty of harnessing some of the real world in tight theories is also known. Certain aspects, notably capturing behavior of natural phenomena (such as weather) or social phenomena (such as people's reactions) can be treated within the framework of such theories only to a limited extent. A major task now is to build on the foundation provided by the successes of fuzzy set theory and AI to create intelligent software which can assist humans in the data-rich environment of the future. This field has been labeled soft computing, computational intelligence, information engineering, etc. We are not interested in the debate that rages between the proponents of these various terms, nor in their precise definitions. There is more important work to be done in building intelligent software systems and refining theory to underpin these new developments. The aim is to produce software which is intelligent in the sense of being able to cope with imprecise and/or uncertain inputs, whilst also able to carry out computations and reasoning with the goal of supporting the human user. This requires more than one theoretical framework -- it requires a paradigm that can be successfully blended with other approaches relevant to the problem domain. Fuzzy logic is in a unique position with respect to this requirement. It can be incorporated into existing problem-solving
vi paradigms to the extent needed, helping therefore to build on existing solutions, and to state and solve more complex problems. This book is based on the third workshop on Fuzzy Logic in AI held in conjunction with IJCAI'95 in Montreal. It contains extended versions of most of the papers presented at the workshop and a number of other invited papers. It is once again a testimony to the fact, well known among fuzzy logic researchers, but little known elsewhere, that fuzzy techniques have something very important to contribute to intelligent systems and their conception, design, and deployment. Some of the latest issues under scrutiny by fuzzy logic researchers are presented. Many researchers have recognized where fuzzy techniques can bring a degree of flexibility and tolerance to noise, enabling effective systems to be developed. There is a need to move away from the flat "if-then" statements of fuzzy control into more general rule-based systems, embodying hierarchical rules and knowledge-based techniques such as casebased reasoning. Alternative extensions arise from the combination of neural technologies with fuzzy systems. The first section of this volume is devoted to hybrid and novel architectures. The next two sections are based on two traditional AI areas, learning and vision. The problems of refining knowledge from raw data and of representing and improving human expert knowledge are particularly suited to a fuzzy approach. This is partly because of the inherent robustness of fuzzy, and partly because fuzzy terms are readily understood by humans. Computer vision is another area of AI where fuzzy techniques can bring a degree of flexibility and tolerance to noise, enabling effective systems to be developed. The final section covers more theoretical areas, including possibility theory and analogical reasoning. In its content, the volume focuses on the most pressing problems of AI. In the approaches presented, it supports the view that fuzzy systems combined with traditional AI will lead the move towards the next generation of intelligent systems. We hope that by providing a snapshot of some current research in the field this book will be both interesting and useful for its readers. November 1996 Trevor Martin, Bristol, UK Anca Ralescu, Cincinnati, USA
Contents Hybrid and Novel Architectures Constructing Prioritized Fuzzy Models R. R. Yager Integra6ng Activities with Nenrofuzzy Distributed Systems A. B. S. Serapiao and A. F. Rocha The Use of Fuzzy Representation in a CBR System for Mesh Design N. Hurley FLIP++ A Fuzzy Lo~c hfferenee Processor Library M. Bonner, S. Mayer, A. Raggl, and W. Slany Fuzzy Reasolfing and Applications for Intelligent Scheduling of Robots E. Levner, L. Meyzin, and A. Ptuskin Fuzzy Lo~c as Interfacing Teclufique in Hybrid AI-Systems C. S. Herrmann 1 14 29 44 57 69 Machine Learning and Data Mining Extracting Kalowledge t'rom Data Using an Intelligent Fuzzy Data Browser 81 J. F. Baldwin and T. P. A tartin Fuzzy Systems with Learning Capability 101 S. Abe Automatic I'~lowledge Base Tuning 116 L. Sztcmdera A Fuzzy-Based Approach to the Awalysis of Financi~d Investanents 128 V. Loia and S. Scandizzo Searching for the Organizational Mcanory Using Fuzzy Modeling 144 A. Cannavacciuolo, G. Capaldo, A. Venire, A. Volpe, and G. Zollo Image Processing and Computer Vision Fuzzy Geodesic Distance in Images 153 I. Bloch Using Fuzzy Information in Knowledge Gnided Segmentation of Brain Tumors 167 M. C. Clark, L. 0. Hall. D. B. Goldgof and M. S. Silbiger FEDGE - Fuzzy Edge Detection by Fuzzy Categoriz,ation and Classification of Edges 182 K. H. L. Ho and N. Ohnishi Towards Hybrid Spatial Reasoning 197 H. W. Guesgen
VIII Mobile Robot Localization Using Fuzzy Maps J. Gasds and A. Marthl Structure Co~tition from Images A. L. Ralescu and J. G. Shanahan 207 225 Theoretical Developments Towards Possibilistic Decision Theory D. Dubois and H. Prade Measurement-Theoretic Fr~uneworks for Fuzzy Set Theory T. Bilgi 9 and I. B. Ti#'ksen A Resemblance Approach to Analogical Reasoning Functions B. Bouchon-Meunier and L. Valverde 240 252 266