Fuzzy Logic Outline Fuzzy logic is a set of mathematical principles for knowledge representation based on the membership Part 1 function. Introduction, or what is fuzzy thinking? Andrew Kusiak Intelligent Systems Laboratory Fuzzy sets Unlike two-valued Boolean logic, fuzzy logic is multi-valued. It deals with the degree of 2139 Seamans Center Linguistic variables and hedges The University of Iowa membership and the degree of truth. Fuzzy logic Iowa City, Iowa 52242 - 1527 Operations of fuzzy sets uses the continuum of logical values between 0 [email protected] Fuzzy rules (completely false) and 1 (completely true). Instead http://www.icaen.uiowa.edu/~ankusiak Summary of just black and white, it employs the spectrum of Tel: 319 - 335 5934 Fax: 319 - 335 5669 colours, accepting that things can be partly true and partly false at the same time. Based on material provided by Professor Michael Negnevitsky Fuzzy sets The classical example in fuzzy sets is tall men. Range of logical values in Boolean and fuzzy logic The elements of the fuzzy set “tall men” are all men, but their degrees of membership depend on their height. The concept of a set is fundamental to mathematics. Degree of Membership Name Height, cm Crisp Fuzzy Chris 208 1 1.00 0 0110 1 0 0 0.2 0.4 0.6 0.81 1 A natural language expresses sets. For example, a Mark 205 1 1.00 (a) Boolean Logic. (b) Multi-valued Logic. car indicates a set of cars. When we say a car, we John 198 1 0.98 Tom 181 1 0.82 mean one out of the set of cars. David 179 0 0.78 Mike 172 0 0.24 Bob 167 0 0.15 Steven 158 0 0.06 Bill 155 0 0.01 Peter 152 0 0.00 Crisp and fuzzy sets of “tall men” A fuzzy set is a set with fuzzy boundaries. Degree of Crisp Sets Membership Let X be the universe of discourse and its elements 1.0 The x-axis represents the universe of discourse − the 0.8 range of all possible values applicable to the be denoted as x. In the classical set theory, crisp Tall Men 0.6 variable of choice. In our case, the variable is the set A of X is defined as function fA(x) called the 0.4 man height. According to this representation, the characteristic function of A 0.2 0.0 universe of men’s heights consists of all tall men. ⎧1, if x∈ A 150160 170 180 190 200 210 f (x) = Height, cm fA(x): X → {0, 1}, where A ⎨ Degree of 0,if x∉ A Membership Fuzzy Sets ⎩ 1.0 The y-axis represents the membership value of the 0.8 fuzzy set. In our case, the fuzzy set of “tall men” This set maps universe X to a set of two elements. 0.6 maps height values into corresponding membership For any element x of universe X, characteristic 0.4 function f (x) is equal to 1 if x is an element of set 0.2 values. A 0.0 A, and is equal to 0 if x is not an element of A. 150160 170 180 190 200 210 Height, cm Crisp and fuzzy sets of short, average, and tall men In the fuzzy theory, fuzzy set A of universe X is Representation of crisp and fuzzy subsets Degree of Crisp Sets defined by function μA(x) called the membership Membership 1.0 μ (x) function of set A X Fuzzy Subset A 0.8 Short Average ShortTall Tall Men 0.6 1 μA(x): X → [0, 1], where μA(x) = 1 if x is totally in A; 0.4 μA(x) = 0 if x is not in A; 0.2 0 < μ (x) < 1 if x is partly in A. 0.0 A 150160 170 180 190 200 210 Height, cm 0 This set allows a continuum of possible choices. Degree of Fuzzy Sets x Membership Crisp Subset A Fuzziness Fuzziness For any element x of universe X, membership 1.0 function μ (x) equals the degree to which x is an 0.8 Typical functions that can be used to represent a fuzzy A Short Average Tall 0.6 set are sigmoid, gaussian and pi. However, these element of set A. This degree, a value between 0 0.4 and 1, represents the degree of membership, also 0.2 Tall functions increase the computation time. Therefore, called membership value, of element x in set A. 0.0 most applications use linear fit functions. 150160 170 180 190 200 210 Linguistic variables and hedges The range of possible values of a linguistic variable In fuzzy expert systems, linguistic variables are used represents the universe of discourse of that variable. in fuzzy rules. For example: For example, the universe of discourse of the The fuzzy set theory is rooted in linguistic IF wind is strong linguistic variable speed might be the range between variables. THEN sailing is good 0 and 220 km/h and may include such fuzzy subsets as very slow, slow, medium, fast, and very fast. IF project_duration is long A linguistic variable carries with it the concept of A linguistic variable is a fuzzy variable. For THEN completion_risk is high example, the statement “John is tall” implies that fuzzy set qualifiers, called hedges. Hedges are terms that modify the shape of fuzzy the linguistic variable John takes the linguistic IF speed is slow sets. They include adverbs such as very, somewhat, value tall. THEN stopping_distance is short quite, more or less and slightly. Representation of hedges in fuzzy logic (2/2) Fuzzy sets with the hedge very Representation of hedges in fuzzy logic (1/2) Mathematical Hedge Graphical Representation Mathematical Degree of Expression Hedge Graphical Representation Membership Expression 1.3 A little [μ ( x )] 1.0 A 4 Short ShortTall Very very [μA ( x )] 0.8 1.7 0.6 Average Slightly [μA ( x )] More or less μA ( x ) 0.4 2 Very Short VeryVery Tall Tall Very [μ ( x )] 0.2 Tall A Somewhat μA ( x ) 0.0 3 150160 170 180 190 200 210 Extremely [μA ( x )] 2 2 [μA ( x )] Height, cm if 0 ≤ μ ≤ 0.5 Indeed A 2 Example, for Alex is tall, μ = .86, Alex is very tall, μ = μ2 =.74, 1 − 2 [1 − μA ( x )] Alex is very very tall, μ = μ4 =.55 if 0.5 < μA ≤ 1 Cantor’s sets Operations of fuzzy sets Complement Not A Crisp Sets: Who does not belong to the set? B Fuzzy Sets: How much do elements not belong to th The classical set theory developed in the late 19 A AA the set? century by Georg Cantor describes interactions The complement of a set is an opposite of this set. between crisp sets. For example, if we have the set of tall men, its Complement Containment complement is the set of NOT tall men. When we These interactions are called operations. remove the tall men set from the universe of discourse, we obtain the complement. If A is the fuzzy set, its complement ¬A can be found as ABB AA B follows: Not A μ¬A(x) = 1 −μA(x) A Intersection Union Intersection Containment Union Crisp Sets: Which element belongs to both sets? Crisp Sets: Which sets belong to which other sets? Crisp Sets: Which element belongs to either set? Fuzzy Sets: To what degree the element is in both sets? Fuzzy Sets: How much of the element is in either set? Fuzzy Sets: Which sets belong to other sets? In classical set theory, an intersection between two sets Similar to a Chinese box, a set can contain other The union of two crisp sets consists of every element contains the elements shared by these sets. For that falls into either set. For example, the union of sets. The smaller set is called the subset. For example, the intersection of the set of tall men and the example, the set of tall men contains all tall men; tall men and fat men contains all men who are tall set of fat men is the area where these sets overlap. In OR fat. In fuzzy sets, the union is the reverse of the very tall men is a subset of tall men. However, the fuzzy sets, an element may partly belong to both sets tall men set is just a subset of the set of men. In intersection. That is, the union is the largest with different memberships. A fuzzy intersection is the membership value of the element in either set. The crisp sets, all elements of a subset entirely belong to lower membership in both sets of each element. The a larger set. In fuzzy sets, however, each element fuzzy operation for forming the union of two fuzzy fuzzy intersection of two fuzzy sets A and B on sets A and B on universe X can be given as: can belong less to the subset than to the larger set. universe of discourse X: Elements of the fuzzy subset have smaller μA∪B(x) = max [μA(x), μB(x)] = μA(x) ∪μB(x), B μA∩B(x) = min [μA(x), μB(x)] = μA(x) ∩μB(x), memberships in it than in the larger set. A B AA where x∈X A B where x∈X Operations of fuzzy sets Fuzzy set operators Fuzzy rules What is a fuzzy rule? μ ( x ) μ ( x ) Cantor’s operators B In 1973, Lotfi Zadeh published his second most A fuzzy rule can be defined as a conditional 1 1 A A influential paper.