Computational Intelligence Application Areas to reproduce these statistical qualities.This area can be extended to many areas of application and represents the ability of a neural network system to be "creative." Control Systems The use of neural networks in control systems is one of the fastest-growing appli- cation areas.It is enjoying widespread implementation for several reasons.First, a neural network-based control system can deal with all of the nonlinearities of a system.(The system doesn't have to be approximated as linear.)Second,a network can be used to model the nonlinear system in the process of designing the con- trol system.Third,the development time for a neural network control system is typically much shorter than it is for other more traditional techniques. The number of specific neural network applications for each of the five areas grows,it seems,daily.Some applications are specific to a discipline.For exam- ple,applications in medicine include EEG waveform classification and appendicitis diagnosis.In business and finance,neural networks are part of systems for trading options on commodity futures contracts and finance company credit application processing.Military-related applications include target tracking and recognition, fault diagnoses in aircraft,and the detection of trace amounts of explosives.In the automotive industry,neural networks can determine the battery pack state-of- charge in an electric vehicle,help determine the proper distance a car should follow another,and,in fact,simultaneously control the positions of a number of cars on an expressway.Artistic endeavors are supported as well,with neural networks that can compose music.Other applications cut across disciplines,such as networks for speech recognition,text-to-speech conversion,and image processing. Evolutionary Computation The two main areas of application for evolutionary algorithms are optimization and classification.Most of the discussion in this text focuses on optimization,since most engineering applications of evolutionary computation are related to optimization. Optimization One of the early applications that popularized genetic algorithms was the control of gas pipeline transmission (Goldberg 1989).Evolutionary algorithms have also been applied to multiple-fault diagnosis,robot track determination,schedule opti- mization,conformal analysis of DNA,load distribution by an electric utility,neural network explanation facilities,and product ingredient mix optimization.(In some of these cases,other CI paradigms have been used,too.) Classification A use of evolutionary computation that has applications across many fields, including both classification and optimization,is the evolution of neural networks. This computational intelligence-based methodology is discussed in detail in
Computational Intelligence Application Areas to reproduce these statistical qualities. This area can be extended to many areas of application and represents the ability of a neural network system to be "creative." Control Systems The use of neural networks in control systems is one of the fastest-growing application areas. It is enjoying widespread implementation for several reasons. First, a neural network-based control system can deal with all of the nonlinearities of a system. (The system doesn't have to be approximated as linear.) Second, a network can be used to model the nonlinear system in the process of designing the control system. Third, the development time for a neural network control system is typically much shorter than it is for other more traditional techniques. The number of specific neural network applications for each of the five areas grows, it seems, daily. Some applications are specific to a discipline. For example, applications in medicine include EEG waveform classification and appendicitis diagnosis. In business and finance, neural networks are part of systems for trading options on commodity futures contracts and finance company credit application processing. Military-related applications include target tracking and recognition, fault diagnoses in aircraft, and the detection of trace amounts of explosives. In the automotive industry, neural networks can determine the battery pack state-ofcharge in an electric vehicle, help determine the proper distance a car should follow another, and, in fact, simultaneously control the positions of a number of cars on an expressway. Artistic endeavors are supported as well, with neural networks that can compose music. Other applications cut across disciplines, such as networks for speech recognition, text-to-speech conversion, and image processing. Evolutionary Computation The two main areas of application for evolutionary algorithms are optimization and classification. Most of the discussion in this text focuses on optimization, since most engineering applications of evolutionary computation are related to optimization. Optimization One of the early applications that popularized genetic algorithms was the control of gas pipeline transmission (Goldberg 1989). Evolutionary algorithms have also been applied to multiple-fault diagnosis, robot track determination, schedule optimization, conformal analysis of DNA, load distribution by an electric utility, neural network explanation facilities, and product ingredient mix optimization. (In some of these cases, other CI paradigms have been used, too.) Classification A use of evolutionary computation that has applications across many fields, including both classification and optimization, is the evolution of neural networks. This computational intelligence-based methodology is discussed in detail in
Chapter One-Foundations Chapter 6.Other classification applications include rule-based machine learning systems,such as that used to learn control of pipeline operations by Goldberg(1989) (which also had an optimization element)and classifier systems for high-level semantic networks. Fuzzy Logic Fuzzy logic is being applied in a wide range of applications in engineering areas ranging from robotics and control to architecture and environmental engineering. Other areas of application include medicine,management,decision analysis,and computer science.As with neural networks,new applications appear almost daily. Two of the major application areas are fuzzy control and fuzzy expert systems. Control Systems Fuzzy control systems have been applied to subway systems,cement kilns,traffic signal systems,home appliances,video cameras,and various subsystems of auto- mobiles including the transmission and brake systems.One application familiar to many is the circuitry inside a video camera that stabilizes the image in spite of the unsteady holding of the camera. Expert Systems Fuzzy expert systems have been applied in the areas of medical diagnostics,for- eign exchange trading,robot navigation,scheduling,automobile diagnostics,and the selection of business strategies,just to name a few.We present an example of the role of fuzzy logic in a scheduling system in Chapter 12. Summary This chapter provides background information from which to learn about CI and its implementation.We introduce the definitions and component methodologies of CI,and we debunk some of the myths you may have heard.Having understood the biological basis for the component methodologies,you will be able to better con- ceptualize how these systems work.Briefly reviewing some application areas offers an idea of the types of problem that computational intelligence tools can be used to solve. Exercises 1.What are some alternative terms for processing element?Discuss the choices, listing advantages and disadvantages for each
Chapter OnemFoundations Chapter 6. Other classification applications include rule-based machine learning systems, such as that used to learn control of pipeline operations by Goldberg (1989) (which also had an optimization element) and classifier systems for high-level semantic networks. Fuzzy Logic Fuzzy logic is being applied in a wide range of applications in engineering areas ranging from robotics and control to architecture and environmental engineering. Other areas of application include medicine, management, decision analysis, and computer science. As with neural networks, new applications appear almost daily. Two of the major application areas are fuzzy control and fuzzy expert systems. Control Systems Fuzzy control systems have been applied to subway systems, cement kilns, traffic signal systems, home appliances, video cameras, and various subsystems of automobiles including the transmission and brake systems. One application familiar to many is the circuitry inside a video camera that stabilizes the image in spite of the unsteady holding of the camera. Expert Systems Fuzzy expert systems have been applied in the areas of medical diagnostics, foreign exchange trading, robot navigation, scheduling, automobile diagnostics, and the selection of business strategies, just to name a few. We present an example of the role of fuzzy logic in a scheduling system in Chapter 12. Summary This chapter provides background information from which to learn about CI and its implementation. We introduce the definitions and component methodologies of CI, and we debunk some of the myths you may have heard. Having understood the biological basis for the component methodologies, you will be able to better conceptualize how these systems work. Briefly reviewing some application areas offers an idea of the types of problem that computational intelligence tools can be used to solve. Exercises .............. 1. What are some alternative terms for processing element? Discuss the choices, listing advantages and disadvantages for each
Exercises 15 2.State a myth relative to neural networks,fuzzy systems,or evolutionary computation,in addition to those discussed in this chapter.Why is it a myth? 3.How do you think adaptation and self-organization are interrelated? 4.Survey recent technical publications and the Internet for these additional areas to which one of the component technologies of CI has been successfully applied: face recognition,health screening,creating art. a.What motivated the use of the technology in these applications? b.What technical tools,in addition to CI,were required to solve the problems? c.What was the role of the CI component technology in each case? 5.What is the difference between fuzziness and probability?Provide an example to illustrate the difference. 6.What is the definition of artificial intelligence?List some differences between computational intelligence and artificial intelligence
Exercises 2. State a myth relative to neural networks, fuzzy systems, or evolutionary computation, in addition to those discussed in this chapter. Why is it a myth? 3. How do you think adaptation and self-organization are interrelated? 4. Survey recent technical publications and the Internet for these additional areas to which one of the component technologies of CI has been successfully applied: face recognition, health screening, creating art. a. What motivated the use of the technology in these applications? b. What technical tools, in addition to CI, were required to solve the problems? c. What was the role of the CI component technology in each case? 5. What is the difference between fuzziness and probability? Provide an example to illustrate the difference. 6. What is the definition of artificial intelligence? List some differences between computational intelligence and artificial intelligence
chapter two Computational Intelligence This chapter covers the key elements of Despite the relatively widespread use of computational intelligence and how com- the term computational intelligence,there putational intelligence fits into the larger is no commonly accepted definition of the picture comprising machine intelligence term.The definitions offered in Chapter 1 and biological intelligence.We examine include assumptions about the nature of adaptation and learning,how they differ, what are called the "constituent method- and what that means for computational ologies"of computational intelligence.As intelligence (CI).We build from the bot-will be seen,other researchers make dif- tom up,identifying each element in turn.ferent assumptions and arrive at different First we discuss three main types of perspectives. adaptation that are incorporated into a As is true for researchers in any develop- variety of computational models:super- ing,maturing field,we are standing on the vised,unsupervised,and reinforcement shoulders of those who have preceded us adaptation.Next we briefly examine the Of particular influence has been work pub concept of self-organization,which we lished by Marks(1993)and Bezdek(1981 believe plays an important role in evo-1992,1994,1998).An extension of their lution.We then look at how computa-work presented in this chapter is a new tional intelligence has been perceived and model of biological and machine intelli- defined by various researchers.Finally,we gence that defines the context for compu- discuss our view of computational intel- tational intelligence. ligence and how it fits into a model of This chapter is not meant to be the intelligent systems. final word on any aspect of computational 17
chapter Computational Intelligence This chapter covers the key elements of computational intelligence and how computational intelligence fits into the larger picture comprising machine intelligence and biological intelligence. We examine adaptation and learning, how they differ, and what that means for computational intelligence (el). We build from the bottom up, identifying each element in turn. First we discuss three main types of adaptation that are incorporated into a variety of computational models: supervised, unsupervised, and reinforcement adaptation. Next we briefly examine the concept of self-organization, which we believe plays an important role in evolution. We then look at how computational intelligence has been perceived and defined by various researchers. Finally, we discuss our view of computational intelligence and how it fits into a model of intelligent systems. Despite the relatively widespread use of the term computational intelligence, there is no commonly accepted definition of the term. The definitions offered in Chapter 1 include assumptions about the nature of what are called the "constituent methodologies" of computational intelligence. As will be seen, other researchers make different assumptions and arrive at different perspectives. As is true for researchers in any developing, maturing field, we are standing on the shoulders of those who have preceded us. Of particular influence has been work published by Marks (1993) and Bezdek (1981, 1992, 1994, 1998). An extension of their work presented in this chapter is a new model of biological and machine intelligence that defines the context for computational intelligence. This chapter is not meant to be the final word on any aspect of computational 17
18 Chapter Two-Computational Intelligence intelligence.It is intended only to be a snapshot in time,and a relatively subjective snapshot at that.If it stimulates discussion and further development,it will accomplish our objective. With those caveats,the chapter is initiated by discussing adaptation and pre- senting several definitions.None of these definitions is meant to be particularly controversial.Rather,they are intended to provide the framework for the remain- der of the book. ■ Adaptation We discuss adaptation and,later,self-organization because they play an important role in our view of computational intelligence.The concept of adaptation is central to computational intelligence.One definition stated in Chapter 1 is that computa- tional intelligence comprises practical adaptation concepts,paradigms,algorithms, and implementations that enable or facilitate appropriate actions(intelligent behav- ior)in complex and changing environments. Webster's New Collegiate Dictionary's(1991)definition of adaptation provides a useful beginning to our discussion: 1:the act or process of adapting:the state of being adapted 2:adjustment to envi- ronmental conditions:as a:adjustment of a sense organ to the intensity or quality of stimulation b:modification of an organism or its parts that makes it more fit for existence under the conditions of its environment. The same source defines the word adapt as follows:"to make fit (as for a spe- cific or new use or situation)often by modification."To be fit is to be suitable, that is,adapted so as to be capable of surviving and acceptable from a particular viewpoint. Thus,we define adaptation as the ability of a system to change,or evolve,its parameters in order to better meet its goal.Dynamic adaptation is the ability of a system to adapt"online,"that is,in essentially real time,in a changing environment. In dynamic adaptation,the system adapts while it is running (online),rather than being taken offline to be retrained.For a system to exhibit adaptation,its trajectory through the problem space must depend on the state of its environment. Accordingly,a number of factors can make adaptation difficult(Holland 1992): 1.A large problem space(the hyperspace comprising the dynamic ranges of all problem variables),which contains many alternative(candidate) solutions,called structures. 2.A large number of variables in each structure,making difficult the deter- mination of which variables,and which combinations of variables, contribute to good solutions
Chapter Two---Computational Intelligence intelligence. It is intended only to be a snapshot in time, and a relatively subjective snapshot at that. If it stimulates discussion and further development, it will accomplish our objective. With those caveats, the chapter is initiated by discussing adaptation and presenting several definitions. None of these definitions is meant to be particularly controversial. Rather, they are intended to provide the framework for the remainder of the book. • Adaptation We discuss adaptation and, later, self-organization because they play an important role in our view of computational intelligence. The concept of adaptation is central to computational intelligence. One definition stated in Chapter 1 is that computational intelligence comprises practical adaptation concepts, paradigms, algorithms, and implementations that enable or facilitate appropriate actions (intelligent behavior) in complex and changing environments. Webster's New Collegiate Dictionary's (1991) definition of adaptation provides a useful beginning to our discussion: 1: the act or process of adapting: the state of being adapted 2: adjustment to environmental conditions: as a: adjustment of a sense organ to the intensity or quality of stimulation b: modification of an organism or its parts that makes it more fit for existence under the conditions of its environment. The same source defines the word adapt as follows: "to make fit (as for a specific or new use or situation) often by modification." To be fit is to be suitable, that is, adapted so as to be capable of surviving and acceptable from a particular viewpoint. Thus, we define adaptation as the ability of a system to change, or evolve, its parameters in order to better meet its goal. Dynamic adaptation is the ability of a system to adapt "online," that is, in essentially real time, in a changing environment. In dynamic adaptation, the system adapts while it is running (online), rather than being taken offline to be retrained. For a system to exhibit adaptation, its trajectory through the problem space must depend on the state of its environment. Accordingly, a number of factors can make adaptation difficult (Holland 1992): 1. A large problem space (the hyperspace comprising the dynamic ranges of all problem variables), which contains many alternative (candidate) solutions, called structures. 2. A large number of variables in each structure, making difficult the determination of which variables, and which combinations of variables, contribute to good solutions