Fuzzy Sets and Systems 141 (2004) 1 – 3 www.elsevier.com/locate/fss
Preface
Genetic fuzzy systems. New developments While preparing our participation in the Joint IFSA-NAFIPS 2001 International Conference, working in the organization of a mini-track on genetic fuzzy systems, we realized that 2001 was the tenth anniversary of the -rst publications on genetic fuzzy systems. The papers by Karr [3], Pham and Karaboga [4], Thrift [5], and Valenzuela-Rend6on [6], were all published in 1991. For that reason, the mini-track included the paper “Ten years of genetic fuzzy systems: Current framework and new trends”, as an attempt to summarize previous research and to open a view to the future of the -eld. In addition, 2001 was also the date when the -rst authored book on the state of the art of genetic fuzzy systems was published [1], 1 almost simultaneously with the celebration of the Joint IFSA-NAFIPS Conference (Vancouver). Considering that date as a sort of milestone, we decided to compile in a special issue (this one) some papers describing the most recent and innovative works in the -eld, as well as to advert attention to open questions and to identify future trends in genetic fuzzy systems. The result was a call for papers for a special issue on “Genetic Fuzzy Systems: New Developments”, receiving 10 submissions, some of them resulting from extended versions of papers presented at the previously mentioned mini-track, while others directly submitted to the special issue. One of the guest editors, plus two additional reviewers (whose work and cooperation we want to acknowledge and thank) carefully reviewed each of the ten manuscripts. The Co-Editors-inChief of Fuzzy Sets and Systems conducted reviews of papers authored or co-authored by any of the guest editors. At the end, only six of the ten submitted contributions were accepted for publication. The six papers on this special issue address three distinct subjects. An introductory paper provides an overview of the current state-of-the art of the -eld and a perspective on its future. The next three address learning of fuzzy classi-cation systems, two of them introducing boosting techniques in the realm of genetic fuzzy systems, and one based on multi-objective genetic selection. The last two papers deal with a hierarchical co-evolutionary structure to evolve fuzzy models, and an application concerning autonomous navigation of mobile robots, respectively. The papers contents are summarized next.
1
We should note that a previous book was written by A. Geyer-Schulz on genetic fuzzy systems in 1995 [2], but it was focused on very speci-c research topics within the area, the design of fuzzy classi-er systems based on the Michigan-style and the derivation of fuzzy rules by genetic programming techniques. c 2003 Elsevier B.V. All rights reserved. 0165-0114/$ - see front matter doi:10.1016/S0165-0114(03)00110-6
2
Preface / Fuzzy Sets and Systems 141 (2004) 1 – 3
The -rst paper, entitled “Ten years of genetic fuzzy systems: current framework and new trends”, by O. Cord6on, et al. provides an account of the past decade of research and development of genetic fuzzy systems with an emphasis on genetic fuzzy rule-based systems. Initially, the paper addresses the main ideas, models and illustrative applications. Next, it critically evaluates genetic fuzzy systems from the point of view of knowledge extraction. New trends, examples of open questions that remain to be addressed, and a collection of key references to access the -eld conclude the paper. The paper “A fast genetic method for inducing descriptive fuzzy models” by L. S6anchez and J. Otero considers extended fuzzy additive models and suggests a technique that can be regarded as the counterpart to boosting fuzzy classi-ers in fuzzy modeling. They claim that, since extended additive models can be estimated by matching pursuit algorithm, one can expect computationally more eFcient algorithms to learn fuzzy rules from data. Actually, they show that a combination of a genetic algorithm and the back-tting process provides a faster learning algorithm than ad hoc schemes. Next, F. HoHmann presents a new boosting algorithm for genetic learning of fuzzy classi-cation rules in the paper “Combining boosting and evolutionary algorithms for learning of fuzzy classi-cation rules”. The iterative learning scheme constructs the fuzzy rule base in an incremental fashion, repeatedly invoking a genetic algorithm that identi-es the single rule classi-er that best matches the current distribution of training instances. The boosting mechanism reduces the weights of those training instances that are correctly classi-ed by a new rule. Rule generation cycles focus on fuzzy rules that account for previously uncovered or misclassi-ed instances. Comparisons with alternative classi-cation approaches show the eHectiveness of the algorithm. The paper “Fuzzy rule selection by multi-objective genetic local search algorithms and rule evaluation measures in data mining”, by H. Ishibuchi and T. Yamamoto, shows how a small number of simple fuzzy rules can be eHectively selected for pattern classi-cation problems with continuous attributes. Their strategy operates in two distinct phases. The -rst phase generates candidate rules through rule evaluation measures whereas the second step selects rules via multi-objective evolutionary algorithms. M. Regatieri Delgado et al. suggest a co-evolutionary approach to support hierarchical collaborative relations between individuals of a population of fuzzy models in “Coevolutionary genetic fuzzy systems: a hierarchical collaborative approach”. Species mean partial solutions organized into four hierarchical levels. Individuals of each level encode membership functions, rules, rule bases and fuzzy systems, respectively. Evaluation is performed via a shared -tness scheme. Applications concerning function approximation and classi-cation illustrate the eHectiveness of the approach. Finally, a novel genetic fuzzy technique for online learning and adaptation of autonomous robotic navigation system is addressed in the paper “Learning and adaptation of an intelligent robot navigator operating in unstructured environment based on a novel online fuzzy-genetic system” co-authored by H. Hagras et al. The emphasis is on learning obstacle avoidance behaviors via delayed reinforcement. They show how such a behavior can be coordinated with other behaviors that receive immediate reinforcement. The navigator is able to dynamically adapt to new environments by updating its rule base. As guest editors, and in addition to the careful work of the reviewers who helped us to assemble this issue, and the authors who answer our call as well, we would like to thank Michael H. Smith and William A. Gruver, General Chairs of the Joint IFSA-NAFIPS 2001 Conference, and Didier
Preface / Fuzzy Sets and Systems 141 (2004) 1 – 3
3
Dubois and Henri Prade, Co-Editors-in-Chief of Fuzzy Sets and Systems, for the opportunity to edit this issue. O. Cord6on F. Herrera Department of Computer Science and Arti cial Intelligence University of Granada 18071 Granada; Spain E-mail address:
[email protected] E-mail address:
[email protected] F. Gomide Department of Computer Engineering and Industrial Automation FEEC; State University of Campinas SP 13083-970, Brazil E-mail address:
[email protected] F. HoHmann Royal Institute of Technology Center for Autonomous Systems S-10044 Stockholm; Sweden E-mail address:
[email protected] L. Magdalena Department of Applied Mathematics; E.T.S.I. Telecomunicaci6on Politechnic University of Madrid; 28040 Madrid; Spain E-mail address:
[email protected] References [1] O. Cord6on, F. Herrera, F. HoHmann, L. Magdalena, Genetic Fuzzy Systems—Evolutionary Tuning and Learning of Fuzzy Knowledge Bases, World Scienti-c, Singapore, 2001. [2] A. Geyer-Schulz, Fuzzy Rule-Based Expert Systems and Genetic Machine Learning, Physica, Heidelberg, 1995. [3] C. Karr, Genetic algorithms for fuzzy controllers, AI Expert 6 (2) (1991) 26–33. [4] D.T. Pham, D. Karaboga, Optimum design of fuzzy logic controllers using genetic algorithms, J. System Eng. 1 (1991) 114–118. [5] P. Thrift, Fuzzy logic synthesis with genetic algorithms, in: Proceedings of Fourth International Conference on Genetic Algorithms (ICGA’91), San Diego, USA, Morgan Kaufmann, Los Altos, CA, 1991, pp. 509–513. [6] M. Valenzuela-Rend6on, The fuzzy classi-er system: a classi-er system for continuously varying variables, in: Proceedings of Fourth International Conference on Genetic Algorithms (ICGA’91), San Diego, USA, Morgan Kaufmann, Los Altos, CA, 1991, pp. 346–353.
