‫البمجية‬

‫مساق الحوسبة ر‬
Soft Computing 8702661
Dr. Labib Arafeh,
Associate Professor
[email protected]

Ch2 SC Fall 2016 1

 Outlines
• Introduction
• Basic Definitions &Terminology
• Set-theoretic Operations
• Membership Function (MF)
• Set-Theoretic Operations
• MF Formulation & Parameterization
• Fuzzy Union, Intersection and
Complement
Ch2 SC Fall 2016 2
 Introduction

NONFUZZY SET FUZZY SET

100 COOL 100 COOL
MEMBERSHIP
(PERCENTAGE)

10 15 20 10 15 20
Air Temperature Air Temperature
Ch2 SC Fall 2016 3
• Sets with fuzzy boundaries
Degree of Membership of “tall men”
No Name Height (cm)
Crisp Fuzzy
1 Fadi 206 1 1
2 Rami 190 1 1
3 Hani 175 0 0.8
4 Sami 160 0 0.7
5 Hadi 155 0 0.4
A = Set of tall people

Crisp set A Fuzzy set A

1.0 1.0
.9
.5 Membership
function

175 185 190 Heights 175 185 190 Heights

Universe of discourse
Ch2 SC Fall 2016 4
• Characteristics of Membership Functions (MFs):
– Subjective (Individual) measures  MFs specified for
the same concept by different persons may vary
considerably (The “sensible number of children in a
family”). This subjectivity comes from individual
differences in perceiving or expressing abstract concepts
– Not probability functions.
MFs “tall” in Asia

.8

.5 “tall” in the US

.1 “tall” in NBA

175 Heights
Ch2 SC Fall 2016 5
 Basic definitions & Terminology
• Classical Set vs Fuzzy set:
– Let X be the universe of discourse and its elements be
denoted as x.
– In the classical set theory, crisp set A of X is defined as
function fA(x) called the characteristic function of A
1, if x  A

f A ( x ) : X  {0,1}, where f A ( x )  

0, if x  A
–In the fuzzy theory, fuzzy set A of universe of discourse X is
defined by function  A (x) (membership function of set A)
 A ( x) : X  [0,1], where  A ( x)  1 if x is totally in A;
 A ( x)  0 if x is not in A;
0   A ( x)  1 if x is partly in A.
Ch2 SC Fall 2016 6
• Formal definition:
If X is a collection of objects denoted generally by x, then
A fuzzy set A in X is expressed as a set of ordered pairs:
A  {( x ,  A ( x ))| x  X }

Membership
Universe or
Fuzzy set function
universe of discourse
(MF)

A fuzzy set is totally characterized by a

membership function (MF). MF maps each
element of X to a membership grade (or value)
between 0 and 1.
Ch2 SC Fall 2016 7
•Fuzzy Sets with Discrete Universes:
– Fuzzy set C = “desirable city to live in”
X = {JER, RAM, HEB} (discrete and non-ordered)
C = {(JER, 0.9), (RAM, 0.8), (HEB, 0.6)}
(subjective membership values!)
– Fuzzy set A = “sensible number of children”
X = {0, 1, 2, 3, 4, 5, 6} (discrete universe)
A = {(0, .1), (1, .3), (2, .7), (3, 1), (4, .6), (5, .2), (6, .1)}
Member function on a Discrete Universe
(subjective 1
Membership Grades

Membership 0.8

0.6

values!) 0.4

0.2

0
0 1 2 3 4 5 6
X = Number of Children

Ch2 SC Fall 2016 8

• Fuzzy Sets with A Continuous Universe:
– Fuzzy set B = “about 50 years old”
X = Set of positive real numbers of possible ages for
human beings (continuous)
B = {(x, µB(x)) | x in X}

Membership Function on a Continous Universe

1.00
0.90

1 0.80
μB(x)  2
membership Grades

0.70
 x  50 
1  
0.60

 10 
0.50
0.40
0.30
0.20
0.10
0.00
0 10 20 30 40 50 60 70 80 90 100
x = Age

Ch2 SC Fall 2016 9

 • A fuzzy set A can be alternatively denoted as follows:

X is discrete
A  μ (x
x i X
A i )/x i

X is continuous A   μ A (x)/x
X

The Summation & integration signs stand for the union of

membership grades; / stands for a marker and does not
imply division.
• Rewriting previous examples (S9 and S10) using the
above equation:
C = 0.9 / JER+ 0.8 / RAM + 0.6 / HEB
A = 0.1/0 + 0.3/1 + 0.7/2 + 1.0/3 + 0.6/4 + 0.2/5 + 0.1/6

Ch2 SC Fall 2016 10

• Linguistic variables: To express the knowledge:
– Hot, Cold;
– High, Low;
Fadi is tall
– Tall, Short;
– Slow, Fast.

• Hedges (operators): To modify the degree truth of

linguistic variables in the sense of:
– Concentration (Very);
– Dilation (Somewhat);
Rami is very tall
– Intensification (Indeed);
– Power (Very Very).
Ch2 SC Fall 2016 11
• Fuzzy Partition:
– Fuzzy partitions of AGE Can be formed by the
linguistic values
– “young”, “middle aged”, and “old”:
Slide20

Ch2 SC Fall 2016 12

– Support(A) = {x  X | A(x) > 0}
= Ᾱ0
– Core(A) = {x  X | A(x) = 1}
= A1
Core is exactly one point

Ch2 SC Fall 2016 13

Normality: core(A)  0 (nonempty)  A is a normal fuzzy
set (That is, A is normal if its MF equals 1 some where,
otherwise it is sub-normal)
Crossover(A) = {x  X | A(x) = 0.5}

Ch2 SC Fall 2016 14

Fuzzy Singelton:
A fuzzy set whose support is a single point in X with
μA(x) = 1
The fuzzy singleton “45 years old”

Ch2 SC Fall 2016 15

• Convexity of Fuzzy Sets: A fuzzy set A is convex if and
only if for any x1 and x2 ε X and λ ε [0, 1],

Fuzzy Logic with

Engineering Applications
By Timothy J. Ross
Ch2 SC Fall 2016 16
• Fuzzy numbers: a fuzzy number A is a fuzzy set in real line
R that satisfies the conditions for normality & convexity;
– Normalization: The maximum membership grade for an
element in a fuzzy set is said to be normalized.
– A set which is not normalized can be made so by
altering all the membership values in proportion so as
to make the largest value 1.
µ NORM(A) (x)=µ A (x) / max(µ A (x)) for x є X
– The following operation converts a sub-normal, non-
empty fuzzy set into its normal version:

Ch2 SC Fall 2016 17

• Bandwidths: for a normal & convex fuzzy set, the
bandwidth or width is the distance between two unique
crossover points:
Width (A) = |x2 – x1|
Where
A(x1) = A(x2) = 0.5

• Symmetry: A fuzzy set A is symmetric if its MF is

symmetric around a certain point x = c, namely,

A(c + x) = A(c – x) for all (x)  X

Ch2 SC Fall 2016 18
• Open left, open right, closed:
The fuzzy set “young” is Open Left
The fuzzy set “old” is Open Right
The fuzzy set “middle age” is Closed

open left fuzzy set A  lim μ A(x)  1 and lim μ A(x)  0

x -  x  

open right fuzzy set A  lim μ A(x)  0 and lim μ A(x)  1

x -  x  

closed fuzzy set A  lim μ A(x)  lim μ A(x)  0

x -  x  
Ch2 SC Fall 2016 19
 Set-Theoretic Operations
• Equality
• Containment
• Union
• Intersection
• Complement
• Normalization
• Algebraic product:
µA•B(x)=µA(x)•µB(x) for all x є X
• Concentration and Dilation

Ch2 SC Fall 2016 20

• Complement:
o Crisp: What doesn’t belong to the set?
o FS: How much do elements not belong to the set?
• Containment:
o Crisp: Which sets belong to which other sets?
o FS: Which sets belong to other sets?
• Intersection:
o Crisp: Which element belongs to both sets?
o FS: How much of the element is in both sets?
• Refer to Table 2.1, PP. 22 shown on the next slide.
Ch2 SC Fall 2016 21
Ch2 SC Fall 2016 22
 • Containment (Subset): FS A is contained in FS B

A  B   A  B
• Complement:
A  X  A  A ( x )  1  A ( x )

Ch2 SC Fall 2016 23

 • Union:
C  A  B  c ( x)  max(  A ( x), B ( x))   A ( x )  B ( x )

• Intersection:

C  A  B  c ( x )  min( A ( x ), B ( x ))  A ( x ) B ( x )

Ch2 SC Fall 2016 24

MF Formulation & Parameterization
MFs of One Dimension:
 Triangular MF: trimf (x; 20, 60, 80)

 x a c  x 
trimf ( x ; a, b , c )  max min ,  , 0
  b  a c  b 
Traingular MF
1

0.8
Membership Grades

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90 100

Ch2 SC Fall 2016 25

 Trapezoidal MF: trapmf (x; 10, 20, 60, 95)

 x a d  x 
trapmf ( x ; a, b , c , d )  max min , 1,  , 0
 b a d c 

Ch2 SC Fall 2016 26

 Gaussian MF: gaussmf (x; 50, 20)

2
1  x c 
  
gaussmf ( x; c,  )  e 2  

Ch2 SC Fall 2016 27

 Generalized bell MF: gbellmf (x; 20, 4, 50)
1
gbellmf ( x; a, b, c) 
xc
2b

1
a

Ch2 SC Fall 2016 28

 – Change of parameters in the generalized bell MF
1
gbellmf ( x; a, b, c) 
xc
2b

1
a

Ch2 SC Fall 2016 29

Physical meaning of parameters in a generalized bell MF
1
gbellmf ( x; a, b, c) 
xc
2b

1
a
Ch2 SC Fall 2016 30
• Gaussian MFs and bell MFs achieve smoothness, they
are unable to specify asymmetric Mfs which are
important in many applications;
• Asymmetric & close MFs can be synthesized using
either the absolute difference or the product of two
sigmoidal functions. sigmf(x; a, c)  1
• Sigmoidal MF: 1  e  a( x c )
Extensions:
Abs. difference
of two sig. MF

Product
of two sig. MF

Ch2 SC Fall 2016 31

• A sigmoidal MF is inherently open right (+ve a), or left
(-ve A) & thus, it is appropriate for representing
concepts such as “very large” or “very negative”;
• Sigmoidal MF are mostly used as activation function of
artificial neural networks (NN);
• A NN should synthesize a close MF in order to simulate
the behavior of a fuzzy inference system;
• Left –Right (L-R) MF:

Ch2 SC Fall 2016 32

 Example: FL ( x)  max( 0, 1  x 2 FR ( x)  max( 0,1  x 2 )

c=65 c=25
a=60 a=10
b=10 b=40

• The list of MFs introduced are not exhaustive;

• Any type of continuous probability distribution
functions can be used as a MF, provided that a set of
parameters is given to specify the appropriate
meanings of the MF.
• Other specialized MFs can be created for specific
applications if necessary;
Ch2 SC Fall 2016 33
• MFs of two dimensions
– In this case, there are two inputs assigned to a MF: this
MF is a two dimensional MF. A one input MF is called
ordinary MF;
– Extension of a one-dimensional MF to a 2-dimensional
MF via cylindrical extensions:
• If A is a fuzzy set in X, then its cylindrical extension in

X x Y is a fuzzy set C(A) defined by: C ( A)   ( x ) /( x, y )
XxY
A

• C(A) can be
viewed as a
2-D fuzzy
set;

Ch2 SC Fall 2016 34

 • Projection of fuzzy sets (decrease dimension)
– Let R be a two-dimensional fuzzy set on X x Y. Then
the projections of R onto X and Y are defined as:
and
respectively.
Two-dimensional Projection Projection
MF onto X onto Y

Ch2 SC Fall 2016 35

 • Composite & non-composite MFs: (MFs of 2-D are
either composite or non-composite.)
– Suppose that the fuzzy A = “(x,y) is near (3,4)” is
defined by:   x  3 2 2
 A (x, y)  exp       y  4  
  2  
  x  3 2    y  4 2 
 exp      exp     
  2     1  
 G(x;3,2) * G( y;4,1)
» This two-dimensional MF is composite
» The fuzzy set A is composed of two statements:
“x is near 3” & “y is near 4”
»These two statements are respectively defined as:
 near 3 (x) = G(x;3,2) and  near 4 (x) = G(y;4,1)
» Expressed as an analytic expression of two 1-D
MFs.
Ch2 SC Fall 2016 36
 » If a fuzzy set is defined by:

» it is non-composite.
• A composite 2-dimensional MF is usually the result of
two statements joined by the AND or OR
connectives.
• Composite 2-dimensional MFs based on min & max
operations
– Let trap(x) = trapezoid (x;-6,-2,2,6)
trap(y) = trapezoid (y;-6,-2,2,6)
be two trapezoidal MFs on X and Y respectively,
Ch2 SC Fall 2016 37
 trap(x) = trapezoid (x;-6,-2,2,6)
trap(y) = trapezoid (y;-6,-2,2,6)
By applying min & max operators, we obtain 2-D MFs
on XxY.

Two dimensional MFs defined by the min and max operators

Min & Max

Operators
for a bell
function
Ch2 SC Fall 2016 38
 More on Fuzzy Union,
Intersection & Complement
– Fuzzy complement
• Another way to define reasonable & consistent
operations on fuzzy sets
– General requirements:
» Boundary: N(0)=1 and N(1) = 0
» Monotonicity: N(a) > N(b) if a < b
» Involution: N(N(a)) = a  Optional
(double complement of a fuzzy set is the set itself)
Ch2 SC Fall 2016 39
 – Two types of fuzzy complements:
» Sugeno’s complement: (s > -1)

(Family of fuzzy complement operators)

» Yager’s complement: (w > 0)

Ch2 SC Fall 2016 40

 – Fuzzy Intersection and Union:

• The intersection of two fuzzy sets A and B is

specified in general by a function

T: [0,1] x [0,1]  [0,1] with ~

 A B (x )  T  A (x ),  B (x )   A (x ) *  B (x )
~
where * is a binary operator for the function T.

This class of fuzzy intersection operators are

called T-norm (triangular) operators.

Ch2 SC Fall 2016 41

•T-norm operators should satisfy Go back to Slide22:
• Boundary: T ( 0 , 0 ) = 0 , T ( a , 1 ) = T ( 1 , a ) = a
Correct generalization to crisp sets

• Monotonicity: T ( a , b ) ≤ T ( c , d ) i f a ≤ c a n d b ≤ d
A decrease of membership in A & B cannot increase a
membership in A  B

• Commutativity: T ( a , b ) = T ( b , a )
T is indifferent to the order of fuzzy sets to be combined

• Associativity: T ( a , T ( b , c ) ) = T ( T ( a , b ) , c )
Intersection is independent of the order of pair wise
groupings
Ch2 SC Fall 2016 42
• T-norm (cont.)

– Four of the most frequently used T-norm operators are:

•Minimum: Tmin(a, b) = min ( a,b) = a  b

•Algebraic product: Tap(a, b) = ab

•Bounded product: Tbp(a, b) = 0 V (a + b – 1)

•Drastic product: Tdp(a, b) =

Ch2 SC Fall 2016 43

Knowing that a & b are between 0 & 1, the surface plots of
these four T-norm operators as functions of a & b are:
Algebraic Bounded Drastic
Minimum: product: product: product:
Tmin(a, b) Tap(a, b) Tbp(a, b) Tdp(a, b)

The corresponding surfaces when

a = μA(x) = a = trapmf(x; 3, 8, 12, 17) and
b = μB(x) = trapmf(y; 3, 8, 12, 17)are:
Minimum: Algebraic Product: Bounded Product: Drastic Product:
Tmin(a, b) Tap(a, b) Tbp(a, b) Tdp(a, b)

Ch2 SC Fall 2016 44

 Both Surfaces together. Observe:
T dp( a , b ) ≤ T bp( a , b ) ≤ T ap( a , b ) ≤ T min( a , b )

Minimum: Bounded product:

Tm(a, b) Tb(a, b)
Algebraic product: Drastic product:
Ta(a, b) Td(a, b)

Ch2 SC Fall 2016 45

• T-conorm or S-norm: operator is a binary mapping S(.,.)
The fuzzy union operator is defined by a function
S: [0,1] x [0,1]  [0,1]
which aggregates two membership function as:
~
 A B  S A (x ),  B (x )   A (x )   B (x )
~
where  is a binary operator for the function s, which
is called an s-norm satisfying:
– Boundary: S ( 1 , 1 ) = 1 , S ( a , 0 ) = S ( 0 , a ) = a
– Monotonicity: S ( a , b ) ≤ S ( c , d ) i f a ≤ c a n d b ≤ d
– Commutativity: S ( a , b ) = S ( b , a )
– Associativity: S ( a , S ( b , c ) ) = S ( S ( a , b ) , c )
Ch2 SC Fall 2016 46
– T-conorm or S-norm (cont.)

•Four T-conorm operators:

–Maximum: S(a, b) = max( a,b) = a V b

–Algebraic sum: S(a, b) = a + b - ab

–Bounded sum: S(a, b) = 1  (a + b)

a, if b  0
–Drastic sum: S(a, b) = 
b, if a  0
1, if a, b  0

Ch2 SC Fall 2016 47
 T-conorm or S-norm. Observe:
Smax(a, b) ≤ Sap(a, b) ≤ Sbp(a, b) ≤ Sdp(a, b)

Maximum: Algebraic sum: Bounded sum: Drastic sum:

S (a, b) S (a, b) S (a, b) S (a, b)

~ ~
 

Ch2 SC Fall 2016 48

 • Generalized DeMorgan’s Law:

– T-norms T(., .) and T-conorms S(., .) are duals which

support the generalization of DeMorgan’s law:

» T(a, b) = N(S(N(a), N(b)))

» S(a, b) = N(T(N(a), N(b)))

N(.) is a fuzzy complement operator

Tm(a, b) Sm(a, b)
Ta(a, b) Sa(a, b)
Tb(a, b) Sb(a, b)
Td(a, b) Sd(a, b)

Ch2 SC Fall 2016 49