Interpretability Issues In Fuzzy Modeling

Fuzzy modeling has become one of the most productive and successful results of fuzzy logic. Among others, it has been applied to knowledge discovery, automatic classification, long-term prediction, or medical and engineering analysis. The research developed in the topic during the last two decades has been mainly focused on exploiting the fuzzy model flexibility to obtain the highest accuracy. This approach usually sets aside the interpretability of the obtained models. However, we should remember the initial philosophy of fuzzy sets theory directed to serve the bridge between the human understanding and the machine processing. In this challenge, the ability of fuzzy models to express the behavior of the real system in a comprehensible manner acquires a great importance. This book collects the works of a group of experts in the field that advocate the interpretability improvements as a mechanism to obtain well balanced fuzzy models.

The key idea of this book is that hinging hyperplanes, neural networks and support vector machines can be transformed into fuzzy models, and interpretability of the resulting rule-based systems can be ensured by special model reduction and visualization techniques. The first part of the book deals with the identification of hinging hyperplane-based regression trees. The next part deals with the validation, visualization and structural reduction of neural networks based on the transformation of the hidden layer of the network into an additive fuzzy rule base system. Finally, based on the analogy of support vector regression and fuzzy models, a three-step model reduction algorithm is proposed to get interpretable fuzzy regression models on the basis of support vector regression. The authors demonstrate real-world use of the algorithms with examples taken from process engineering, and they support the text with downloadable Matlab code. The book is suitable for researchers, graduate students and practitioners in the areas of computational intelligence and machine learning.

Fuzzy logic has found applications in an incredibly wide range of areas in the relatively wide range of areas in the relatively short time since its conception. It was invented by Lotfi Zadeh, a leading systems expert, so it is perhaps not surprising that system theory is one of the areas in which fuzzy logic has made a profound impact. Fuzzy logic combined with the paradigm of computing with words allows the use and manipulation of human knowledge and reasoning in the modeling and control of dynamical systems. This monograph presents new approaches to the construction of fuzzy models and to the design of fuzzy controllers. The emphasis is on developing methods that allow systematic design on the one hand and mathematical analysis of the resulting system on the other. In particular, the methods described allow rigorous analysis of the stability and robustness of the systems, which are crucial issues in control theory. The first theme of the book is a new approach to the system design and analysis of fuzzy controllers, given linguistic information concerning the plant and the control objective. The new approach, fuzzy Lyapunov synthesis, is a computing-with-words version of the well-known (classical) Lyapunov synthesis method. The second theme of the book is to show that fuzzy controllers are in fact solutions to a nonlinear optimal control problem. The authors formulate a novel nonlinear optimal control problem, consisting of a new state-space model -- referred to as the hyperbolic state-space model -- and a new cost functional and show that its solution is a fuzzy controller. This leads to a new framework for fuzzy modeling and control that combines the advantages of the fuzzyworld, such as linguistic interpretability, and of classical optimal control theory, such as guaranteed stability and robustness.

This book gathers cutting-edge papers in the area of Computational Intelligence, presented by specialists, and covering all major trends in the research community in order to provide readers with a rich primer. It presents an overview of various soft computing topics and approximate reasoning-based approaches, both from theoretical and applied perspectives. Numerous topics are covered: fundamentals aspects of fuzzy sets theory, reasoning approaches (interpolative, analogical, similarity-based), decision and optimization theory, fuzzy databases, soft machine learning, summarization, interpretability and XAI. Moreover, several application-based papers are included, e.g. on image processing, semantic web and intelligent tutoring systems. This book is dedicated to Bernadette Bouchon-Meunier in honor of her achievements in Computational Intelligence, which, throughout her career, have included profuse and diverse collaborations, both thematically and geographically.

This book shows that the term “interpretability” goes far beyond the concept of readability of a fuzzy set and fuzzy rules. It focuses on novel and precise operators of aggregation, inference, and defuzzification leading to flexible Mamdani-type and logical-type systems that can achieve the required accuracy using a less complex rule base. The individual chapters describe various aspects of interpretability, including appropriate selection of the structure of a fuzzy system, focusing on improving the interpretability of fuzzy systems designed using both gradient-learning and evolutionary algorithms. It also demonstrates how to eliminate various system components, such as inputs, rules and fuzzy sets, whose reduction does not adversely affect system accuracy. It illustrates the performance of the developed algorithms and methods with commonly used benchmarks. The book provides valuable tools for possible applications in many fields including expert systems, automatic control and robotics.

From theory to techniques, the first all-in-one resource for EIS There is a clear demand in advanced process industries, defense, and Internet and communication (VoIP) applications for intelligent yet adaptive/evolving systems. Evolving Intelligent Systems is the first self- contained volume that covers this newly established concept in its entirety, from a systematic methodology to case studies to industrial applications. Featuring chapters written by leading world experts, it addresses the progress, trends, and major achievements in this emerging research field, with a strong emphasis on the balance between novel theoretical results and solutions and practical real-life applications. Explains the following fundamental approaches for developing evolving intelligent systems (EIS): the Hierarchical Prioritized Structure the Participatory Learning Paradigm the Evolving Takagi-Sugeno fuzzy systems (eTS+) the evolving clustering algorithm that stems from the well-known Gustafson-Kessel offline clustering algorithm Emphasizes the importance and increased interest in online processing of data streams Outlines the general strategy of using the fuzzy dynamic clustering as a foundation for evolvable information granulation Presents a methodology for developing robust and interpretable evolving fuzzy rule-based systems Introduces an integrated approach to incremental (real-time) feature extraction and classification Proposes a study on the stability of evolving neuro-fuzzy recurrent networks Details methodologies for evolving clustering and classification Reveals different applications of EIS to address real problems in areas of: evolving inferential sensors in chemical and petrochemical industry learning and recognition in robotics Features downloadable software resources Evolving Intelligent Systems is the one-stop reference guide for both theoretical and practical issues for computer scientists, engineers, researchers, applied mathematicians, machine learning and data mining experts, graduate students, and professionals.

Fuzzy modeling usually comes with two contradictory requirements: interpretability, which is the capability to express the real system behavior in a comprehensible way, and accuracy, which is the capability to faithfully represent the real system. In this framework, one of the most important areas is linguistic fuzzy modeling, where the legibility of the obtained model is the main objective. This task is usually developed by means of linguistic (Mamdani) fuzzy rule-based systems. An active research area is oriented towards the use of new techniques and structures to extend the classical, rigid linguistic fuzzy modeling with the main aim of increasing its precision degree. Traditionally, this accuracy improvement has been carried out without considering the corresponding interpretability loss. Currently, new trends have been proposed trying to preserve the linguistic fuzzy model description power during the optimization process. Written by leading experts in the field, this volume collects some representative researcher that pursue this approach.

The analysis and control of complex systems have been the main motivation for the emergence of fuzzy set theory since its inception. It is also a major research field where many applications, especially industrial ones, have made fuzzy logic famous. This unique handbook is devoted to an extensive, organized, and up-to-date presentation of fuzzy systems engineering methods. The book includes detailed material and extensive bibliographies, written by leading experts in the field, on topics such as: Use of fuzzy logic in various control systems. Fuzzy rule-based modeling and its universal approximation properties. Learning and tuning techniques for fuzzy models, using neural networks and genetic algorithms. Fuzzy control methods, including issues such as stability analysis and design techniques, as well as the relationship with traditional linear control. Fuzzy sets relation to the study of chaotic systems, and the fuzzy extension of set-valued approaches to systems modeling through the use of differential inclusions. Fuzzy Systems: Modeling and Control is part of The Handbooks of Fuzzy Sets Series. The series provides a complete picture of contemporary fuzzy set theory and its applications. This volume is a key reference for systems engineers and scientists seeking a guide to the vast amount of literature in fuzzy logic modeling and control.