Skip to main content
Fachgebiete
Automobil + Motoren Bauwesen + Immobilien Business IT + Informatik Elektrotechnik + Elektronik Energie + Nachhaltigkeit Finance + Banking Management + Führung Marketing + Vertrieb Maschinenbau + Werkstoffe Versicherung + Risiko
DE
EN
Themenseiten
Organisationspsychologie Projektmanagement Marketing Smart Manufacturing
Bücher
Zeitschriften
Weitere Formate
Podcasts Webinare Technik Webinare Wirtschaft Kongresse Veranstaltungskalender Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen
Referenzkunden
MTZ-Fachkongress

Antriebe und Energiesysteme von morgen


Austausch von Automobil- und Energiewirtschaft

Am 14./15. Mai geht es in Chemnitz um die Mobilitätswende durch Elektro- und Wasserstofffahrzeuge.
Erweiterte Suche
Anmelden
nach oben

2002 | Buch

Introduction Kapitel lesen Erstes Kapitel lesen

Neuro-Fuzzy Architectures and Hybrid Learning

verfasst von: Professor Danuta Rutkowska

Verlag: Physica-Verlag HD

Buchreihe : Studies in Fuzziness and Soft Computing

Enthalten in: Professional Book Archive

Einloggen, um Zugang zu erhalten
insite
SUCHEN

Über dieses Buch

The advent of the computer age has set in motion a profound shift in our perception of science -its structure, its aims and its evolution. Traditionally, the principal domains of science were, and are, considered to be mathe­ matics, physics, chemistry, biology, astronomy and related disciplines. But today, and to an increasing extent, scientific progress is being driven by a quest for machine intelligence - for systems which possess a high MIQ (Machine IQ) and can perform a wide variety of physical and mental tasks with minimal human intervention. The role model for intelligent systems is the human mind. The influ­ ence of the human mind as a role model is clearly visible in the methodolo­ gies which have emerged, mainly during the past two decades, for the con­ ception, design and utilization of intelligent systems. At the center of these methodologies are fuzzy logic (FL); neurocomputing (NC); evolutionary computing (EC); probabilistic computing (PC); chaotic computing (CC); and machine learning (ML). Collectively, these methodologies constitute what is called soft computing (SC). In this perspective, soft computing is basically a coalition of methodologies which collectively provide a body of concepts and techniques for automation of reasoning and decision-making in an environment of imprecision, uncertainty and partial truth.

Inhaltsverzeichnis

Frontmatter
1. Introduction
Abstract
The initial idea behind writing this book was to present the new neuro-fuzzy architectures and the novel hybrid learning algorithms, developed as results of research into implication-based neuro-fuzzy systems and learning methods. These research projects were conducted in the Department of Computer Engineering, Technical University of Czestochowa, Poland, and supervised by the author of this book. Some of the conclusions have been published in papers contributed by the author, as well as Dr. Robert Nowicki and Dr. Artur Starczewski, who are the author’s former Ph.D. students. Their Ph.D. dissertations [366], [479] were prepared on the basis of the above mentioned research.
Danuta Rutkowska
2. Description of Fuzzy Inference Systems
Abstract
Approximate reasoning, based on fuzzy sets and fuzzy logic, has been successfully employed in fuzzy inference systems. These systems are used in many practical applications, mainly as fuzzy controllers, but also as other knowledge-based systems such as expert systems, fuzzy classifiers and so on. Fuzzy systems have been recently combined with neural networks and genetic algorithms to create different kinds of neuro-fuzzy systems and intelligent systems. This chapter presents an overview of fuzzy sets, approximate reasoning, and fuzzy systems.
Danuta Rutkowska
3. Neural Networks and Neuro-Fuzzy Systems
Abstract
This chapter presents an overview of neural networks and neuro-fuzzy systems. The latter are a fusion of neural networks and fuzzy techniques, introduced in [293], initially developed in [66], [408], [87], and then in [167], [166], [491], [273], [169], [157], [228], [503], [270], and others. Neuro-fuzzy systems have been applied in many consumer products [492], [493]. They incorporate some merits of both neural networks and fuzzy systems. In the neuro-fuzzy combinations we distinguish fuzzy neural networks (see Section 3.2), obtained by introducing fuzziness directly into neural networks [169], and fuzzy inference neural networks (see Section 3.3), which are representations of fuzzy systems in the form of connectionist networks [513], similar to neural networks. Of course, different types of neuro-fuzzy systems can be found in the literature, e.g. [493], [300], [53], [162], [361], [243], [347], [582], [229], [223], [496], [244], [56], [141], [101].
Danuta Rutkowska
4. Neuro-Fuzzy Architectures Based on the Mamdani Approach
Abstract
The fuzzy inference neural networks (see Section 3.3) that realize the inference based on the Mamdani approach are the subject of this chapter. Different, multi-layer, architectures of the neuro-fuzzy systems are portrayed. The systems with various fuzzifiers (singleton, non-singleton), defuzzifiers, and inference operations, are considered. All these systems can be trained, when applied to solve practical problems, similarly to neural networks. Learning methods of neuro-fuzzy systems are presented in Chapter 6, including the architecture-based learning, proposed in Section 6.1.3. Interested readers may also be referred to [420], [434].
Danuta Rutkowska
5. Neuro-Fuzzy Architectures Based on the Logical Approach
Abstract
The fuzzy inference neural networks (see Section 3.3) that realize the in-ference based on the logical approach are the subject of this chapter. Firstly,in Section 5.1, the mathematical descriptions of the neuro-fuzzy systemsemploying different fuzzy implications are determined. Then, the connec-tionist, multi-layer, architectures, which correspond to the implicationbased systems, are presented. These architectures are proposed in [366].The neuro-fuzzy systems of this kind are considered in [367], [430], [433], and also in the papers that refer to a specific implication, e.g. [429]. In Section 5.4, the performance analysis of the implication-based systems is illustrated. The results of computer simulations with regard to examples of function approximation, control, and classification problems, are portrayed in Section 5.5. In order to train the systems, gradient, genetic, or hybrid al-gorithms can be applied. The learning methods are described in Chapter 6. In particular, the architecture-based learning, outlined in Section 6.1.3, is recommended.
Danuta Rutkowska
6. Hybrid Learning Methods
Abstract
In Chapters 4 and 5 the connectionist, multi-layer architectures of fuzzy systems, called fuzzy inference neural networks, were presented. These architectures are similar to neural networks (see Section 3.1), so learning algorithms can be proposed to tune the parameters of the networks, analo gously to tuning weights in neural networks. The parameters of the neuro fuzzy architectures define the shape of membership functions of the fuzzy sets in the IF-THEN rules. Tuning these parameters thus optimizes the form of the rules. Moreover, the number of rules in the rule base of the fuzzy systems can be determined using a learning method. The number of elements in the first layers of the neuro-fuzzy architectures depends on the number of the rules, so this kind of algorithms determines the architectures. Hybrid learning, which is the subject of this chapter, consists of a combina tion of different learning methods, such as gradient, genetic, and clustering algorithms. These methods are first described, and then the hybrid algo rithms for rule generation and parameter tuning are presented, including the algorithms proposed in [479], [438], [439], [440], [480], [481].
Danuta Rutkowska
7. Intelligent Systems
Abstract
The neurofuzzy architectures and hybrid learning procedures, described in the previous chapters, can be employed to create so-called intelligent com putational systems. A general schema of these kind of systems is presented in this chapter. Intelligent systems usually refer to the field of Artificial In telligence (AT) or Computational Intelligence (CI). The difference between these branches of Computer Science is explained in Section 7.1. Then, ex pert systems are outlined (Section 7.2). Intelligent computational systems (Section 7.3) can be viewed as a special type of expert systems. Finally, in Section 7.4, perception-based systems are considered as intelligent systems in AI.
Danuta Rutkowska
8. Summary
Abstract
DIt should be noted that the “pure” fuzzy system, depicted in Fig. 2.19, is a special case of perception-based systems, considered in Section 7.4. Since this system, with a fuzzifier and a defuzzifier, can be represented in the connectionist form of a neuro-fuzzy system, it seems that perception-based systems, or at least some types of them, may also be combined with neural networks to create perception-based neuro-fuzzy systems. These systems might be trained using different hybrid learning methods. Having learning ability makes these systems even more “intelligent.” It would be interesting to combine these systems with fuzzy neural networks (see Section 3.2), and include genetic (evolutionary) algorithms to support the learning process. Fuzzy neural networks that process granular information refer to granular neural networks [388]. Networks of this kind may be applied with reference to perception-based systems, since the latter also use information granulation
Danuta Rutkowska
Backmatter
Metadaten
Titel
Neuro-Fuzzy Architectures and Hybrid Learning
verfasst von
Professor Danuta Rutkowska
Copyright-Jahr
2002
Verlag
Physica-Verlag HD
Electronic ISBN
978-3-7908-1802-4
Print ISBN
978-3-7908-2500-8
DOI
https://doi.org/10.1007/978-3-7908-1802-4