<p> Computer Engineering Department *** February 2, 2000</p><p>BASED ON EXTRACTS FROM PUBLICATIONS OF L.ZADEH, B.KOSKO, G.KLIR, BO YUAN, U.ST.CLAIR, M.JAMSHIDI, R.BERKAN, S.TRUBATCH, E.COX, LI-XIN WANG, A.KANDEL, W.PEDRYCZ, R.PACHERO, A.MARTINS, S.KHATOR </p><p>In the literature sources, we can find different kinds of justification for fuzzy systems theory. Human knowledge nowadays becomes increasingly important – we gain it from experiencing the world within which we live and use our ability to reason to create order in the mass of information (i.e. to formulate human knowledge in a systematic manner). Since we are all limited in our ability to perceive the world and to profound reasoning, we find ourselves everywhere confronted by uncertainty which is a result of lack of information (lexical impression, incompleteness), in particular, inaccuracy of measurements. The other limitation factor in our desire for precision is a natural language used for describing/sharing knowledge, communication, etc. We understand core meanings of word and are able to communicate accurately to an acceptable degree, but generally we cannot precisely agree among ourselves on the single word or terms of common sense meaning. In short, natural languages are vague.</p><p>Our perception of the real world is pervaded by concepts which do not have sharply defined boundaries – for example, many, tall, much larger than, young, etc. are true only to some degree and they are false to some degree as well. These concepts (facts) can be called fuzzy or gray (vague) concepts – a human brain works with them, while computers may not do it (they reason with strings of 0’s and 1’s). Natural languages, which are much higher in level than programming languages, are fuzzy whereas programming languages are not. The door to the development of fuzzy computers was opened in 1985 by the design of the first logic chip by Masaki Togai and Hiroyuki Watanabe at Bell Telephone Laboratories. In the years to come fuzzy computers will employ both fuzzy hardware and fuzzy software, and they will be much closer in structure to the human brain than the present-day computers are.</p><p>In 1965 Lotfi A.Zadeh of the University of California, Berkeley, published his paper “Fuzzy Sets” [Information and Control, vol.8, pp.338-353] where the word fuzzy meant vague in the technical literature. The dictionary defines vague as lacking precision or sharpness of definition, indistinct. A statement of fact like “He is tall” does not have a binary truth value. It has a vague of “fuzzy” truth value (degree of truth) in the range between 0 and 1. Thus, fuzziness can be considered as vagueness of description of semantic meaning of statements or events. Vague concepts (or, fuzzy sets) like high temperature depend on fuzzy degrees of truth or set membership. </p><p>The concept of a fuzzy set contrasts with a classical concept of a bivalent set (crisp set), whose boundary is required to be precise, i.e. a crisp set is a collection of things for which it is known whether any given thing is inside it or not. L.Zadeh generalized the idea of a crisp set by extending a valuation set {1,0} (definitely in/definitely out) to the interval of real values (degrees of membership) between 1 and 0 denoted as [0,1]. We can say that the degree of membership of any particular element of a fuzzy set expresses the degree of compatibility of the element with a concept represented by fuzzy set. It means that a fuzzy set A contains an object x to degree a(x), i.e. a(x) = Degree(xA), and the map a : X  {Membership Degrees} is called a set function or MEMBERSHIP FUNCTION. The fuzzy set A can be expressed as A = {(x, a(x)) }, xX, and it imposes an elastic constrain of the possible values of elements xX called the possibility distribution. Fuzzy sets tend to </p><p>1 Computer Engineering Department *** February 2, 2000</p><p> capture vagueness exclusively via membership functions that are mappings from a given universe of discourse X to a unit interval containing membership values. A membership function can be either discrete or continuos. The notation of the fuzzy set in each case is the following:</p><p>[discrete case] A = X a(x)/x a ( x ) / x [continuos case] A = X </p><p>(both symbols  and  stand for the union of all elements in the set).</p><p>In each application of fuzzy set theory, we must construct appropriate fuzzy sets (i.e. their membership functions) which adequately capture the intended meanings of relevant linguistic terms. The problem of constructing membership functions is not a problem of fuzzy set theory per se. It is a problem of knowledge acquisition, which is a subject of a relatively new field called KNOWLEDGE ENGINEERING. Knowledge acquisition extracts useful and reproducible knowledge from (1) experts, (2) data sets, (3) textbook information and (4) commonsense reasoning applied to a specific objective. Besides, the knowledge engineer is also involved in the process – his role is to elicit the knowledge of interest from the experts and to express it in some operational form of a required type. </p><p>In its broad sense, the term “FUZZY LOGIC” means an application area of fuzzy set theory (its concepts, principles and methods) for formulating various forms of sound APPROXIMATE REASONING. The latter (often-called FUZZY REASONING) can be understood as the process of inferring imprecise conclusions from imprecise premises. This deduction is “approximate” rather than “exact” because data and implications1 are described by vague concepts. </p><p>In practice fuzzy logic means computing with words. Since computation with words is possible, computerized systems can be built by embedding human expertise articulated in daily language. Also called a FUZZY INFERENCE ENGINE or FUZZY RULE-BASE, such a system can perform approximate reasoning somewhat similar to (but much more primitive) than that of the human brain. Computing with words seems to be a slightly futuristic phrase today since only certain aspects of natural language can be represented by the calculus of fuzzy sets, but still fuzzy logic remains one of the most practical ways to mimick human expertise in a realistic manner. The fuzzy approach uses a premise that humans don’t represent classes of objects (e.g. “class of bald men”, or the “class of numbers which are much greater than 50”) as fully disjoint but rather as sets in which there may be grades of membership intermediate between full membership and non-membership. Thus, a fuzzy set works as a concept that makes it possible to treat fuzziness in a quantitative manner.</p><p>Fuzzy sets form the building blocks for fuzzy IF-THEN rules which have the general form “IF X is A THEN Y is B”, where A and B are fuzzy sets. The term “FUZZY SYSTEMS” refers mostly to systems that are governed by fuzzy IF-THEN rules. The IF part of an implication is called the antecedent whereas the second, THEN part is a consequent.</p><p>A FUZZY SYSTEM is a set of fuzzy rules that converts inputs to outputs. The basic configuration of a pure fuzzy system is shown in Figure 1. The fuzzy inference engine (algorithm) combines fuzzy IF-THEN rules into a mapping from fuzzy sets in the input space X to fuzzy sets in the output space Y based on fuzzy logic principles. From a knowledge representation viewpoint, a fuzzy IF-THEN rule is a scheme for capturing knowledge that involves imprecision. The main feature of reasoning using these rules is its partial matching capability, which enables an inference to be made from a fuzzy rule even when the rule’s condition is only partially satisfied. </p><p>1 Implications or conditional propositions are usually encountered in English in the form of “if…then” rules (propositions)</p><p>2 Computer Engineering Department *** February 2, 2000</p><p>Fuzzy Rule Base</p><p> fuzzy sets in X Fuzzy Inference Engine fuzzy sets in Y</p><p>Figure 1. Basic configuration of a pure fuzzy system</p><p>Generalizing what was mentioned above, we can conclude that a fuzzy set represents (models) fuzziness (vague concepts can be modeled using the notion of partial membership) as well as uncertainty (fuzzy set represents the state of knowledge of a property of an event or object). Fuzziness is a measure of how well an instance (value) conforms to a semantic ideal or concept. For example, in the fuzzy set HOT, the value of temperature 28°C has a degree 0.76 meaning that it is sufficiently (relatively) compatible with HOT. At the same time, value 50°C (as a temperature of air) has a minimal fuzziness or ambiguity (degree of uncertainty is equal or close to zero), and it can be classified as highly compatible with concept HOT.</p><p>It is important to observe that there is an intimate connection between FUZZINESS and COMPLEXITY. As the complexity of a task (problem), or of a system for performing that task, exceeds a certain threshold, the system must necessarily become fuzzy in nature. L.Zadeh, originally an engineer and systems scientist, was concerned with the rapid decline in information afforded by traditional mathematical models as the complexity of the target system increased. As he stressed, with the increasing of complexity our ability to make precise and yet significant statements about its behavior diminishes. Real-world problems (situations) are too complex, and the complexity involves the degree of uncertainty – as uncertainty increases, so does the complexity of the problem. Traditional system modeling and analysis techniques are too precise for such problems (systems), and in order to make complexity less daunting we introduce appropriate simplifications, assumptions, etc. (i.e. degree of uncertainty or FUZZINESS) to achieve a satisfactory compromise between the information we have and the amount of uncertainty we are willing to accept. In this aspect, fuzzy systems theory is similar to other engineering theories, because almost all of them characterize the real world in an approximate manner.</p><p>By making frequent, simplifying assumptions, the problem has become too uncertain to be of practical use as well. In this view of modeling complex systems, the underlying processes (mechanics) can be represented linguistically (with the best utilization of available knowledge) rather than mathematically to treat uncertainties in </p><p>3 Computer Engineering Department *** February 2, 2000</p><p> the system. The fuzzy system uses commonsense rules in place of the mathematical model or so-called plant model. It means that a shift is done towards HUMAN’S REASONING based not on discrete symbols and numbers, but on fuzzy sets. The latter allows mathematical representations to become compatible with expressions in natural languages. In some cases, fuzzy technology makes a solution possible that would otherwise be unthinkable due to cost of computing, whereas by selecting the number of fuzzy representative sets, there is a way of adjusting (or, controlling) the precision level of a solution. Typically, a fuzzy system uses human friendly commands, embodies expert knowledge, yields robust performance, requires reasonable computation time and effort, and allows precision/cost adjustments.</p><p>Fuzzy systems, on one hand, are rule-based systems that are constructed from a collection of linguistic rules; on the other hand, fuzzy systems are NONLINEAR MAPPINGS of inputs (stimuli) to outputs (responses), i.e. certain types of fuzzy systems can be written as compact nonlinear formulas. The inputs and outputs can be numbers or vectors of numbers. These rule-based systems can in theory model any system with arbitrary accuracy, i.e. they work as UNIVERSAL APPROXIMATORS.</p><p>The Achilles’ heel of a fuzzy system is its rules; smart rules give smart systems and other rules give smart systems and other rules give less smart or even dumb systems. The number of rules increases exponentially with the dimension of the input space (number of system variables). This rule explosion is called the CURSE OF DIMENSIONALITY and is a general problem for mathematical models. For the last 5 years several approaches based on decomposition (cluster) merging and fusing have been proposed to overcome this problem. </p><p></p><p>This text was prepared on the base of the following sources:  L.A.Zadeh. The Birth and Evolution of Fuzzy Logic // Int. J. General Systems, vol.17, 1990, pp.95-105  L.A.Zadeh. Fuzzy Sets and Systems // Int. J. General Systems, vol.17, 1990, pp.129-138  G.J.Klir, U.H.St.Clair, Bo Yuan. Fuzzy Sets Theory. Foundations and Applications, Prentice Hall PTR, 1997, ISBN 0-13-341058-7  B.Kosko. Fuzzy Engineering, Prentice Hall, 1997, ISBN 0-13-353731-5  A.Kandel, R.Pachero, A.Martins, S.Khator. Foundations of Rule-Based Computations in Fuzzy Models / Fuzzy Modeling. Paradigms and Practice, W.Pedrycz (ed.), Kluwer Academic Publ., 1996, pp.232-260  Li-Xin Wang. A Course in Fuzzy Systems and Control, Prentice Hall, 1997, ISBN 0-13-593005-7  M.Jamshidi. Large-Scale Systems : Modeling, Control, and Fuzzy Logic, Prentice Hall PTR (Prentice Hall Series on Environmental and Intelligent Manufacturing Systems), 1997, ISBN 0-13-125683-1  R.C.Berkan, S.L.Trubatch. Fuzzy Systems Design Principles. Building Fuzzy IF-THEN Rule Bases, IEEE Press, 1997, ISBN 0-7803-1151-5  E.Cox. The Fuzzy Systems Handbook, Academic Press, CA, 1994  W.Pedrycz. Shadowed Sets: Representing and Processing Fuzzy Sets // IEEE Transactions on Systems, Man and Cybernetics, Part B: Cybernetics, vol.28, #1, 1998 </p><p>4 Computer Engineering Department *** February 2, 2000</p><p> Copyright belongs to the respected authors</p><p>5</p>