Browse Books

Reviewer: Herbert Toth

Soft computing is a new paradigm, a partnership of fuzzy logic, neurocomputing, and probabilistic reasoning, and this is one of the first books on the subject. Probabilistic reasoning comprises belief networks and stochastic, gradient-free optimization techniques such as genetic algorithms and simulated annealing. According to the preface, “The book is intended for use as a text in courses on computational intelligence at either the senior or first-year graduate level. It is also suitable for use as a self-study guide.…Prerequisites are minimal; the reader is expected to have basic knowledge of elementary calculus and linear algebra” (p. xix) as well as probability theory. The text is organized into an introduction and seven parts, as follows. Part 1, “Fuzzy Set Theory,” introduces the reader to the basic definitions of fuzzy set theory, the concepts of fuzzy rules and fuzzy reasoning, and fuzzy inference systems as used in fuzzy control. Part 2, “Regression and Optimization,” presents least-squares methods for system identification. It then describes derivative-based nonlinear techniques and derivative-free optimization techniques (such as genetic algorithms and simulated annealing). Part 3, “Neural Networks,” gives an overview of a number of important neural network paradigms, such as adaptive networks and supervised, reinforcement, and unsupervised learning networks. Part 4, “Neuro-Fuzzy Modeling,” presents Adaptive Neuro Fuzzy Inference Systems (ANFIS) and Coactive ANFIS (CANFIS), two concepts developed by the authors, together with learning rules and application examples. Part 5, “Advanced Neuro-Fuzzy Modeling,” covers some structure identification techniques for neural networks and fuzzy modeling: the Classification and Regression Tree method (CART), fuzzy c-means clustering, and mountain clustering. It also presents techniques for rule-based structure identification. Part 6, “Neuro-Fuzzy Control,” discusses various approaches to the design of neuro-fuzzy controllers and summarizes the pros and cons of each of these methods. Part 7, “Advanced Applications,” presents a number of application examples in different domains, such as printed character recognition, channel equalization, adaptive noise cancellation, and handwritten numeral recognition. It also discusses the use of soft computing in game playing and color recipe prediction. The book ends with four appendices—“Hints to Selected Exercises,” “List of Internet Resources,” “List of MATLAB Programs,” and “List of Acronyms”—and an index. The text is well structured, well illustrated, and easily intelligible. Each chapter starts with an introduction, providing a short overview of the rest of the chapter. All of the chapters in Parts 1 through 6 close with summaries and some exercises. (A solution manual is available from the publisher.) References are given at the end of each chapter. Throughout the book, readers are provided with a fine balance between the mathematical background and the intuitive explanations, as well as between the theoretical concepts and their application possibilities. Thus, I can recommend the book as a comprehensive source of information on neuro-fuzzy and soft computing.