MEEM 1713

ARTIFICIAL INTELLIGENCE AND APPLICATIONS

Chapter 4:
Introduction to Fuzzy Logic
Part 4 – Fuzzy Logic and Fuzzy Set Operations

•Assoc. Prof. Ts. Dr. Herman Wahid

•Department of Control & Mechatronics Engineering (CMED)
•Faculty of Electrical Engineering
•Universiti Teknologi Malaysia.
 Chapter Outline

4.1 Overview of Fuzzy Systems

4.2 Definition of Fuzzy Sets

4.3 Fuzzy Set Operations

4.4 Fuzzy Relations

4.5 Prelude to Fuzzy Logic Control

•www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 2
 4.1
Overview of fuzzy systems
 What is fuzzy logic?

•www.utm.my
 Concept of fuzzy logic control system
 History of Fuzzy Logic
 Applications of Fuzzy Systems

MEEM 1713 – ARTIFICIAL INTELLIGENCE

3
 Introduction
• What is Fuzziness?
According to the Oxford English Dictionary,
The word “fuzzy” means blurred, fluffy, frayed or indistinct.

In Malay, fuzzy is “kabur” or “samar” or “kurang jelas”.

Some important points to note:

 Fuzziness is deterministic uncertainty.
 Fuzziness is connected with the degree to which events occur

www.utm.my
rather than the likelihood of their occurrence (probability).
 For example, the degree to which a person is young is a fuzzy
event rather than a random event.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 4

 How Does Fuzzy Logic
Resembles Human Intelligence?
 Rule Base System (reflects some forms of the Human
Reasoning Process)
 Handling certain levels of Imprecision and Uncertainty
 Clustering and Classification

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 5
 Example of Differences Between
Fuzzy Logic and Crisp Logic

Precise Imprecise
Properties Properties

Full Membership Partial Membership

Yes or No Yes --> No
True or False True --> False
1 or 0 1 ---> 0

Crisp Sets Fuzzy Sets

www.utm.my
Girls age 13 yrs Girls about 13yrs
People 1.5m tall People about 1.5m tall

Crisp Models Fuzzy Models

Crisp relations Fuzzy relations
MEEM 1713 – ARTIFICIAL INTELLIGENCE 6
 Features of Fuzzy Logic

Example of Linguistic Imprecision

Unusual and Real-Life Quotes
How was the weather like yesterday?
Oh! It was rainy with 98% humidity and hot with temperature of 35.5 °C
Oh! It was very humid and really hot.

www.utm.my
Fuzzy logic can handle such linguistic imprecision
where other techniques have difficulty in handling

MEEM 1713 – ARTIFICIAL INTELLIGENCE 7

 Fuzzy logic can handle linguistic imprecision
where other techniques have difficulty in handling

www.utm.my
When you are at 10 metres from the junction start braking at 50% pedal
level.

When you are near the junction, start braking slowly.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 8

 Example of Basic Fuzzy Set Operations
Equality:
A=B <=> mA = mB
Complement
mA~ ( x ) = 1 − mA ( x )

Intersection, AND
mA∩ B ( x ) = min{mA ( x ), mB ( x )}
Union, OR

www.utm.my
m A∪ B ( x) = max{m A ( x), mB ( x)}

The next module will introduce more fuzzy set operations

MEEM 1713 – ARTIFICIAL INTELLIGENCE 9
 Among Many Applications Fuzzy
Logic is Most Successful in Control

Fuzzy Controller

Plant

www.utm.my
In this course much of the applications will be
centred around fuzzy logic control systems.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 10

 A Fuzzy Logic Controller
Consists of 4 Main Parts
2

Defuzzification
Inferencing

Fuzzification
Rule Base 3
If X is A
And Y is B
1 4
Then Z is C

www.utm.my
Sensor Control
Readings Signals

MEEM 1713 – ARTIFICIAL INTELLIGENCE 11

 Why Fuzzy Logic?
Fuzzy Logic can:
• Represent vague language
naturally
• Enrich not replace crisp
sets
• Allow flexible engineering
design
• Improve model
performance

www.utm.my
• Are simple to implement
• And best of all they often
work!
MEEM 1713 – ARTIFICIAL INTELLIGENCE 12
 Brief History of Fuzzy Logic

• 1965 - Seminal Paper by Prof. Lotfi Zadeh on Fuzzy Sets

• 1966 - Fuzzy Logic (P. Marinos, Bell Labs)

• 1972 - Fuzzy Measure (M. Sugeno, TIT)

• 1974 - Fuzzy Logic Control (E.H. Mamdani, London Q. Mary)

• 1980 - Control of Cement Kiln (F.L. Smidt, Denmark)

• 1987 - Sendai Subway Train Experiment, Japan (Hitachi)

• 1988 - Stock Trading Expert System (Yamaichi Securities)

www.utm.my
• 1989 - LIFE (Lab for International Fuzzy Eng.)

Japanese Govt. provides US$70million on Fuzzy research

MEEM 1713 – ARTIFICIAL INTELLIGENCE 13

 Brief History of Fuzzy Logic (Contd.)

1989 - First Fuzzy Logic Air-Conditioner

1990 - First Fuzzy Logic Washing Machine
1990~1994 - Japanese Companies develops
Fuzzy Logic application in a big way

1994 - Japanese companies sold over

US$34billion of fuzzy logic consumer products

1992 ~ 1998 - Research on neuro-fuzzy

www.utm.my
techniques
2000 – now – More applications

MEEM 1713 – ARTIFICIAL INTELLIGENCE 14

 The Success of Fuzzy Logic is mainly due
its introduction into Consumer Products

Early Applications started in Japan since 1980s, some examples are:

 Sendai Subway System
 Temperature Controlled in Showers
 Air Conditioner

www.utm.my
Tokyo’s Electric Town
MEEM 1713 – ARTIFICIAL INTELLIGENCE 15
 The major success of Fuzzy Logic in the mid-eighties is mainly
due to its introduction into Consumer Products

Some Examples are: In 1990 Fuzzy Logic Consumer

Products entered Japanese
Consumer Market in a Big Way:
- Washing Machines
- Camcorder
- Refrigerators
- Televisions

www.utm.my
- Rice Cookers
- Air Conditioners
- Brake control of vehicles
- Heaters
MEEM 1713 – ARTIFICIAL INTELLIGENCE 16
 In 1990, with the introduction of Fuzzy Logic Washing
Machines, extremely high sales were recorded!

www.utm.my
This prompted many other consumer product manufacturers
to start using fuzzy logic in their products.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 17

 Example of fuzzy rules in washing machines

If transmittance is low and load is

heavy then extend washing cycle
longer

If transmittance is high and load

is heavy then extend washing
Photo cycle medium
sensor

Load Sensor

www.utm.my
Light Receptor

Drained out
water
MEEM 1713 – ARTIFICIAL INTELLIGENCE 18
 Fuzzy Logic Automatic
Gear Shift Control in Vehicles

Sensor 1: Road Winding Conditions

Output: Gear
Fuzzy Logic
Shift Pattern

Sensor 2: Frequency of Braking

www.utm.my
Fuzzy logic is used to determine the
best gear position in difficult conditions

MEEM 1713 – ARTIFICIAL INTELLIGENCE 19

 Fuzzy Logic Automatic Gear Shift Control
Example of the Rules

 IF Road is Too_Winding AND Brake is applied

Many_Times THEN Gear is Low

 If Road is Normal AND No Brake is Applied THEN

Gear is Normal

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 20
 Supervisory Elevator Control by Fuzzy Logic

Hall Call
Group Controller States of Each Car
Car Position
Command Car Calls
Direction etc.

Car Controller 1 Car Controller 2 Car Controller n

Car Call
Actuator Actuator Actuator

www.utm.my
Elevator 1 Elevator 2 Elevator n
MEEM 1713 – ARTIFICIAL INTELLIGENCE 21
 Introduction to Fuzzy Logic

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 22
 4.2
Definition of Fuzzy Sets

•www.utm.my
 Fuzzy sets compared with crisp set
 Fuzzy set notations
 Membership functions

MEEM 1713 – ARTIFICIAL INTELLIGENCE

23
 Fuzzy Sets: An Overview

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 24
 Definition of Fuzzy Sets

4.2.1 Fuzzy Sets compared with Crisp sets

A ‘crisp’ set, A, can be defined as a set

which consists of elements with either full µA
or no membership at all in the set. Crisp set, A
1.0

Each item in its universe is either in the

set, or not.

www.utm.my
0
u

MEEM 1713 – ARTIFICIAL INTELLIGENCE 25

 Fuzzy Sets
• Fuzzy sets allow the elements in its set to have partial membership.
• Membership from 0 to 1.
• Thus, a fuzzy set is a generalization of an ordinary set by allowing a
degree (or grade) of membership for each element.
• A fuzzy set uses a common function such as triangular, trapezoidal,
gaussian, etc. called membership functions to support the membership
value of its elements.
µA

Membership value
Fuzzy set A

www.utm.my
1.0

Membership function

Universe of discourse u
MEEM 1713 – ARTIFICIAL INTELLIGENCE 26
 4.2.2 Fuzzy Set Notations

U is called the universe of discourse and u represents a generic element of U.

A fuzzy set F in a universe of discourse U is characterized by a membership
function,
μ F : U → [0, 1]

A fuzzy set F in U is usually represented as a set of ordered pairs of elements

u and grade of membership value:

www.utm.my
F = {(u, µ F (u )) u ∈ U }

MEEM 1713 – ARTIFICIAL INTELLIGENCE 27

 • When U is continuous, a fuzzy set F can be written concisely as:
F = ∫ µ (u ) / u
u
F

• When U is discrete, a fuzzy set F is represented as:

F = Σµ F (ui ) / ui
or

F = µ F (u I ) / u I + µ F (u 2 ) / u 2 + ... + µ F (u N ) / u N

www.utm.my
or

𝐹𝐹 = 𝑢𝑢1 , 𝜇𝜇𝐹𝐹 , 𝑢𝑢2 , 𝜇𝜇𝐹𝐹 , . . . . . .

MEEM 1713 – ARTIFICIAL INTELLIGENCE 28

 Universe of Discourse

• Defined as “a collection of objects all having the same

characteristics”
• Notation: U or X, and elements in the universe of
discourse are: u or x
• Some examples:
• Voltages of actuators
• Speed of cars

www.utm.my
• Error

MEEM 1713 – ARTIFICIAL INTELLIGENCE 29

 Example of Universe of Discourse for
Persons Height

Membership Value, µΗ

Short Medium Tall

1

www.utm.my
140 150 HEIGHT,
160 170 180
H(cm)

MEEM 1713 – ARTIFICIAL INTELLIGENCE 30

 Example 2.1

In the universe of discourse, the fuzzy set F labeled ‘integer

approximately equal to 5’ may be defined as:

F= 0.1/2+0.4/3+0.85/4+1.0/5+0.85/6+0.4/7+0.1/8

Similarly, the fuzzy subset F labeled ‘integer close to 4’ may be

defined as:

F=0.4/2+0.8/3+1/4+0.8/5+0.4/6+0.1/7+0.0/8

www.utm.my
As discussed, fuzzy set (F) can be written in the following form:

F={(2,0.4), (3,0.8), (4,1), (5,0.8), (6,0.4), (7,0.1), (8,0)}

MEEM 1713 – ARTIFICIAL INTELLIGENCE 31

 Definition of Fuzzy Sets
µF
Support set Support set
The support set of a fuzzy set F is the crisp 1.0

set of all points u in U such that

µ F (u ) > 0 u
µA
Crossover point
Crossover point
The element u in U at which is called 1.0
the crossover point when
0.5
µ F (u ) = 0.5
u

www.utm.my
Fuzzy singleton µA
A fuzzy singleton
A fuzzy set whose support is a single point
in U is referred to as a fuzzy singleton.

MEEM 1713 – ARTIFICIAL INTELLIGENCE u 32

 Definition of Fuzzy Sets

µA
α -cut set of a fuzzy set. α-cut set
The α -cut set of a fuzzy set F, labeled with F α
by convention, is the crisp set of all points u in U 1
such that.
µF (u ) ≥ α
α
Note that an α-set discards the point whose
0
membership values are lower than α. u

Normalization µA
This process simply involves re-scaling the NORM(A)
membership function so that its maximum

www.utm.my
1
value is 1, that is: A

µ NORM ( A ) (u ) = µ A (u ) / max(µ A (u )) ∀u ∈ U
0
u
MEEM 1713 – ARTIFICIAL INTELLIGENCE 33
 4.2.3 Membership Functions

Two ways to define the membership for fuzzy sets: numerical and
functional.

Example of a numerical definition.

F= 0.1/2+0.4/3+0.85/4+1.0/5+0.85/6+0.4/7+0.1/8

Example of a functional definition:

www.utm.my
1
f ( x) =
1 + ( x − 5) 2

MEEM 1713 – ARTIFICIAL INTELLIGENCE 34

 We may use standard functions to represent fuzzy sets.

The membership functions which are often used in practice

include:
 s-function.
 π-function.
 triangular form.
 trapezoid form.
 exponential form.

www.utm.my
 gaussian.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 35

 Membership Functions
S-membership function µA
0 for u < a 1.0
2[(u − a )/ (b − a )]2 for a ≤ u ≤ b

S (u; a, b, c ) = 
 1 − 2[(u − c ) / (c − b )]2
for b ≤ u ≤ c 0.5

1 for u > c

a b c u

Π - membership function µA
1.0

www.utm.my
S (u; c-b, c-b/ 2 , c ) for u ≤ c
π (u; b, c ) =  0.5
b
1-S (u; c, c + b/ 2 , c + b ) for u ≥ c
c-b c-b/2 c c+b/2 c+b u

MEEM 1713 – ARTIFICIAL INTELLIGENCE 36

 Membership Functions
Trapezoidal membership function
µ
 0 for u <α
(u − α ) / (β − α ) for α ≤ u < β 1

Π (u;α , β , γ , δ ) =  1 for β ≤ u ≤ γ
 (δ − u ) / (δ − γ ) for γ < u ≤ δ
 0 U
 0 for u >δ α β γ δ

Triangular membership function

µA
0 for u < a

www.utm.my
1.0
(u − a ) / (b − a ) for a ≤ u ≤ b

T (u; a, b, c ) = 
(c − u ) / (c − b ) for b ≤ u ≤ c
0 for u > c
a b c u
MEEM 1713 – ARTIFICIAL INTELLIGENCE 37
 Exercise 4.1

(1). Write down the mathematical 1

expression for the L-membership

0 U
α β
function shown? L-membership function

1
(2). Write down the mathematical

www.utm.my
expression of this membership
function.
-1 0 2 4 5

MEEM 1713 – ARTIFICIAL INTELLIGENCE 38

 Example 4.2
Let the universe of discourse be the interval [0,100] with Y interpreted as
age. A fuzzy set F of Y labeled ‘middle age’ may be defined as :

F = {( y, µ F ( y )) y ∈ y}

What are the implications of the term ‘middle-age?

A person aged 45 is often considered to be middle-aged. However, what

about someone of 38 say? What about other ages? It seems plausible that

•www.utm.my
the required fuzzy set should have a maximum at around 45, and go to zero
below 30 and above 60. Usually, a Gaussian-function or T-function gives a
good description in such circumstances.

MEEM 1713 – ARTIFICIAL INTELLIGENCE

39
 The fuzzy set F might be defined using a T-function as :

0 for 0 ≤ y ≤ 30
( y − 30 ) / 15 for 30 ≤ y ≤ 45

µF (y) = 
(60 − y ) / 15 for 45 ≤ y ≤ 60
0 for 60 ≤ y ≤ 120

1.0

www.utm.my
30 45 60

MEEM 1713 – ARTIFICIAL INTELLIGENCE 40

 Exercise 4.2

Suppose a membership function to show “young” is described by the

following equation:

 1.0 for 0 < x ≤ 25


m( x ) =  1
for x > 25
1 + ( x −525 )2

Based on this membership function, write down the membership values

of the ages for 10, 20, 25, 30, 35, 40, 50 in this set.

www.utm.my
Draw the graph for this membership function to show “youngness”.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 41

 Youngness

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 42
 4.3
Fuzzy Sets Operations
 Basic fuzzy set operations
 Fuzzy set operations for modifying

•www.utm.my
membership functions
 Some advanced fuzzy set operations

MEEM 1713 – ARTIFICIAL INTELLIGENCE

43
 Mapping Sets to Functions

Set-Theoretic Forms  Function-Theoretic Forms.

f: X Y can be defined by:

1, x ∈ A
χ A ( x) = 
0, x ∉ A
where χA expresses membership in set A for the
element x in the universe.

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 44
 Operations for Function-theoretic Terms
Consider 2 sets: A and B, on the universe X.
Union: A ∪ B → χ A∪ B ( x) = χ A ∨ χ B = max(χ A , χ B )

Intersection: A ∩ B → χ A∩ B ( x) = χ A ∧ χ B = min(χ A , χ B )

Complement: A → χ A ( x) = 1 − χ A ( x)

www.utm.my
Containment: A ⊆ B → χ A ( x) ≤ χ B ( x)
MEEM 1713 – ARTIFICIAL INTELLIGENCE 45
 4.3.1 Basic Fuzzy Set Operations

• The use of fuzzy sets provides a basis for the systematic

manipulation of vague and imprecise concepts using fuzzy set
operations performed by manipulating the membership functions.
• Let A and B be two point-valued fuzzy sets in U with membership
functions µA and µB, respectively as given in the next page.
• Consider these two fuzzy sets:
How do you
µA µB perform an
A B operation of

www.utm.my
one fuzzy set
to another?
u u

MEEM 1713 – ARTIFICIAL INTELLIGENCE 46

 Fuzzy Set Operations

Equality
Two fuzzy sets A and B are equal if they are defined on the same
universe and the membership function is the same for both, that is:

A= B iff µ A (u ) = µ B (u ) ∀ u ∈U

Union
The union of two fuzzy sets A and B is the fuzzy set whose
membership function is given by:

www.utm.my
A B
µ A∪ B (u ) = max{µ A (u ), µ B (u )} ∀ u ∈ U

MEEM 1713 – ARTIFICIAL INTELLIGENCE 47

 Fuzzy Set Operations

Intersection
The intersection of two fuzzy sets A and B is the fuzzy set whose
membership function is given by:

µ A  B (u ) = min{µ A (u ), µ B (u )} ∀ u ∈ U

www.utm.my
A B

MEEM 1713 – ARTIFICIAL INTELLIGENCE 48

 Fuzzy Set Operations

Subset
A ⊆ B iff µ A (u ) ≤ µ B (u ) ∀u ∈ U
Draw the graph showing “subset”

Complement
The complement of a (normalized) fuzzy set A with membership
function µA is defined as the fuzzy set on the same universe with

www.utm.my
membership function:
µ A ' (u ) = 1 − µ A (u ) ∀ u ∈ U Not A

MEEM 1713 – ARTIFICIAL INTELLIGENCE 49

 4.3.2 Fuzzy Set Operations
for Modifying Membership Functions

• The shape of membership functions of fuzzy sets can be changed.

• This can be done by using specific operations.

Power
If p is a positive number and A is a fuzzy set with membership
function µA(u), then A on power p, denoted by Ap, is defined as:

www.utm.my
{
A p = {( x, µ A ( x))} = ( x, ( µ A ( x)) p
p
}

MEEM 1713 – ARTIFICIAL INTELLIGENCE 50

 Concentration
A fuzzy set A can be ‘concentrated’ by modifying its membership
function so as to accentuate the membership of the higher
membership elements.
µ CON ( A ) (u ) = (µ A (u )) 2
∀ u ∈U
µA

1
A
CON(A)

www.utm.my
0

Note: This operation is unique with fuzzy sets only. If this operation is carried out
with a crisp set, it results in no change.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 51

 Dilation
A fuzzy set A can be ‘dilated’ by modifying its membership function to
increase the importance of lower membership elements.

µ DIL ( A ) (u ) = (µ A (u ))0.5 ∀ u ∈ U

µA

1
DIL(A)
A

www.utm.my
0

Note: These operations are unique with fuzzy sets only. If they are carried out
with crisp sets, they result in no change.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 52

 Intensification
This operation moves the normalized fuzzy set closer to being crisp, by
enhancing the membership value of those elements whose
membership was above 0.5 and diminishing that of those elements
with membership below 0.5.

2(µ A (u ))2 for 0 ≤ µ A (u ) ≤ 0.5

µ INT ( A ) (U ) = 
1 − 2(1 − µ A (u ))2
for 0.5 ≤ µ A (u ) ≤ 1

µA

www.utm.my
1
A INT(A)

MEEM 1713 – ARTIFICIAL INTELLIGENCE 53

 Exercise 4.3

Consider the fuzzy set A:

A={(1,0.1), (2,0.8), (3,1), (4,0.2), (5, 0.5)}
Solve:
 A3
 CON(A)
 DIL(A)
 INT(A)
Draw graphs showing the operations

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 54
 Solution 4.3

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 55
 Exercise 4.4

For the fuzzy set with the following membership function

(take the elements of the universe of discourse to be 0, 0.5,
1, 1.5 and 2), find and draw the graphs of
CON(A)
DIL(A)

 u for 0 ≤ u ≤ 1
µ A (u ) = 

www.utm.my
2 − u for 1 < u ≤ 2

MEEM 1713 – ARTIFICIAL INTELLIGENCE 56

 Solution 4.4

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 57
 4.3.3 Some Advanced Fuzzy Set Operations

Algebraic product
The algebraic product of two fuzzy sets A and B is the multiplication of
the membership functions which is given by:

µ A⋅B (u ) = µ A (u )µ B (u ) ∀ u ∈U

Algebraic sum
The algebraic sum of two fuzzy sets A and B is also known as

www.utm.my
probabilistic sum and is given by the following expression:
µ A+ˆ B (u ) = µ A (u ) + µ B (u ) - µ A (u )µ B (u ) ∀ u ∈U

MEEM 1713 – ARTIFICIAL INTELLIGENCE 58

 Exercise 4.5

Consider the fuzzy sets A and B where:

A={(x1, 0.2), (x2, 0.4), (x3, 1), (x4, 0.3), (x5, 0.1)}
B={(x1, 0.1), (x2, 0.3), (x3, 0.6), (x4, 0.4)}

Solve:
• algebraic product

www.utm.my
• algebraic sum.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 59

 Bounded product
The bounded product of two fuzzy sets A and B with membership
functions is the fuzzy set whose membership function is given by:

µ AB (u ) = max(0, µ A (u ) + µ B (u ) − 1) ∀u ∈U

Bounded sum
The bounded sum of two fuzzy sets A and B with membership functions
and is the fuzzy set whose membership function is given by:

www.utm.my
µ A⊕ B (u ) = min{1, µ A (u ) + µ B (u ) } ∀ u ∈U

where ‘+’ is the arithmetic sum operator.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 60

 Drastic product

The drastic product of two fuzzy sets A and B with membership function
and is the fuzzy set whose membership function is given by:

µ A (u ) for µ B (u ) = 1

µ A⊗ B (u ) = µ B (u ) for µ A (u ) = 1
0 for µ A (u ), µ B (u ) < 1


www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 61
 Exercise 4.6:
Consider the fuzzy sets A and B where:

A={(x1, 0.2), (x2, 0.3), (x3, 1), (x4, 0.5), (x5, 0.1), }
B={(x1, 0.0), (x2, 0.3), (x3, 0.6), (x4, 0.7), (x5, 0.9)}

Perform the following on A and B:

(1) Algebraic product
(2) Algebraic sum
(3) Bounded product

www.utm.my
(4) Bounded sum
(5) Drastic product

MEEM 1713 – ARTIFICIAL INTELLIGENCE 62

 Table of Summary of Fuzzy Set Operators
Fuzzy set operations Operator expressions

Equality μ A (u ) = μ B (u ) u ∈U
Union μ A B (u ) = max{μ A (u ),μ B (u )} for all u ∈ U
Intersection μ A B (u ) = min{μ A (u ),μ B (u )} for all u ∈ U
Complement μ A (u ) = 1 − μ A (u ) u ∈U
Normalization μ NORM ( A ) (u ) = μ A (u ) / max(μ A (u )) u ∈U
μ CON ( A ) (u ) = (μ A (u )) u ∈U
2
Concentration

μ DIL ( A ) (u ) = (μ A (u )) u ∈U
0.5
Dilation

2(u A (u )) for 0 ≤ µ A (u ) ≤ 0.5

2
Intensification u INT ( A ) (u ) =
1 − 2(1 − u A (u )) for 0.5 ≤ u A (u ) ≤ 1
2

Algebraic product u A• B (u ) = u A (u ) • u B (u )} for all u ∈ U

Bounded sum u A⊕ B (u ) = min{1, u A (u ) + u B (u )} for all u ∈ U
A ⊕ B (u ) = max{0, u A (u ) + u B (u ) − 1} for all u ∈ U

www.utm.my
Bounded product u AB
u A (u ) for u B (u ) = 1

Drastic product u A⊗ B (u ) = u B (u ) for u A (u ) = 1
0 for u A (u ), u B (u ) < 1
•Algebraic Sum 
•Algebraic Sum µ A+ˆ B (u ) = µ A (u ) + µ B (u ) - µ A (u )µ B (u ) ∀ u ∈U
MEEM 1713 – ARTIFICIAL INTELLIGENCE 63
 4.4
Fuzzy Relations
 Fuzzy Relations

•www.utm.my
 Compositional Operators

MEEM 1713 – ARTIFICIAL INTELLIGENCE

64
 2.4.1 Fuzzy Relations
In fuzzy relations, the strength of an r-ary relation between ordered pairs
of r universes is measured with a membership function, expressing
various degrees of strength of the relation on the unit interval [0,1],
µR(x1, x2,…, xn)∈[0,1]

Defining Fuzzy Relations using Cartesian Product

The cartesian product between two fuzzy sets A and B results in a fuzzy
relation R which is contained within the full Cartesian product space:

www.utm.my
A×B=R ⊆ X×Y
where the fuzzy relation R has membership function
µR(x,y)= µA×B(x,y)=min{µA(x), µB(y)}

MEEM 1713 – ARTIFICIAL INTELLIGENCE 65

 Example:

Given two fuzzy sets A (ambient temperature) and B (near optimum

pressure), determine the relation R, defined on the Cartesian product
of the universes of A and B.

 0.2 0.5 1   0.3 0.9 

A= , , , B =  , 
 x1 x2 x3   y1 y2 

0.2 0.2

www.utm.my
R = 0.3 0.5
0.3 0.9

MEEM 1713 – ARTIFICIAL INTELLIGENCE 66

 Example 4.3
Person-Sport Relation (Fuzzy relation)

Consider 2 sets as follows:

A = {Ahmad, Bala, Chong}
B = {Football, Hockey, Badminton, Tennis}

The relation does not simply state either ‘yes’(1) or ‘no’(0) but a more
natural degree of association to each pair is given.
For example a relation matrix as follows is given:
Football
1.0
F H B T Ahmad 0.8

www.utm.my
Hockey
Ahmad 0 0.8 0.7 0 Bala 0.7 0.9
Bala 1.0 0 0.3 0 Badminton
Chong 0.3
Chong 0 0.9 0 0.6 Tennis
0.6

A fuzzy graph showing the (Person, Sport) relation

MEEM 1713 – ARTIFICIAL INTELLIGENCE 67

 Operations on Fuzzy Relations
Suppose R and S are fuzzy relations on Cartesian space X x Y, then the
following operations apply (similar to operations on Fuzzy sets):

Union: μ R ∪S ( x, y ) = max{µ R (x, y ), µ S (x, y )}

Intersection: μ R ∩S ( x, y ) = min{µ R ( x, y ), µ S ( x, y )}

Complement: μ R ( x, y ) = 1 − µ R ( x, y )

www.utm.my
Containment: R ⊂ S ⇒ μ R ( x, y ) ≤ µS ( x, y )

MEEM 1713 – ARTIFICIAL INTELLIGENCE 68

 4.4.2 Compositional Operators

Let the relation R:X→Y and the relation S:Y →Z share a

common universe Y.
The relation T:X → Z can be expressed in terms of R and S
using the composition operation, denoted by,

T=R o S
“o” is the compositional operator.

www.utm.my
This operator is “composed” of at least 2 basic fuzzy set
operations.
Some examples are: max-min, max-max, min-max, etc.
MEEM 1713 – ARTIFICIAL INTELLIGENCE 69
 Popular Compositional Operators

Max-min operation
T=R°S •In fuzzy logic control
application, Max-min and
µT(x,z) = Max {Min [µR(x,y), µS(y,z)]} Max-product composition
operations are the most
Max-product operation commonly and frequently
T=R°S used because of their
computational simplicity
µ T (x,z) = Max [µR(x,y) . µS(y,z)] and efficiency.

Others:

www.utm.my
Max-bounded sum; Max-bounded product;
Max-drastic product; Max-algebraic sum;
Max-Max; Min-Max; etc.
MEEM 1713 – ARTIFICIAL INTELLIGENCE 70
 Example 4.4

Max-min based compositional operation

The composition operator plays an important role in establishing the fuzzy

relation equation for a fuzzy knowledge base. Basically, the composition
operation is the net effect of applying one relation after another. It is defined in
most application, using max-min as follows.

Let P and Q be defined as :

 0.1 0.5   0.3 0.6 0.8
P=  Q= 
0.6 0.9  0.7 0.5 0.4 

www.utm.my
Then the relation, R, can be obtained using the max-min compositional
operations:
 0.1 0.5   0.3 0.6 0.8  R 11 R 12 R 13 
R = PQ =   °  = 
 0.6 0.9   0.7 0.5 0.4   21
R R 22 R 23 

MEEM 1713 – ARTIFICIAL INTELLIGENCE 71

 where:

R11 = max{min (0.1,0.3), min (0.5,0.7 )}

= max{0.1,0.5}
= 0.5

R12 = max{min (0.1,0.6 ), min (0.5,0.5)}

= max{0.1,0.5}
= 0.5

R13 = max{min (0.1,0.8), min (0.5,0.4 )}

= max{0.1,0.4}
= 0.4
R21 = max{min (0.6,0.3), min (0.9,0.7 )}
= max{0.3,0.7}
= 0.7
R22 = max{min (0.6,0.6 ), min (0.9,0.5)}
= max{0.6,0.5}

www.utm.my
= 0.6

R23 = max{min (0.6,0.8), min (0.9,0.4 )}

= max{0.6,0.4}
= 0.6

MEEM 1713 – ARTIFICIAL INTELLIGENCE 72

 Example 4.5

Assume in a university, for the field of AI, there are four courses
offered:
1. Fuzzy Theory (FT).
2. Fuzzy Control (FC).
3. Neural Networks (NN).
4. Expert Systems (ES).
Three students are planning to take these course based on their
different preferences for theory, application, hardware and
programming. Each of these courses contains parts of the four
properties as described.

www.utm.my
The students are: 1. Ali.
2. Bala.
3. Chin.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 73

 Objective of Exercise:

Show that the composition of fuzzy relations can help these students in
their decision making.

There are three different sets: X, Y, and Z. Let.

Set of Students: X = {Ali, Bala, Chin}.
Set of Properties: Y = {Th, Ap, Hw, Pg}
= {theory, application, hardware, prog}.
Set of Courses: Z = {FT, FC, NN, ES}.
= {fuzzy theory, fuzzy control, neural
networks and expert system}.

www.utm.my
Develop the max-min composition of the two fuzzy relations of P(X,Y)
and Q(Y,Z), where this can help the students to choose the proper
courses.
Also solve the problem using Max-product composition.
MEEM 1713 – ARTIFICIAL INTELLIGENCE 74
 Based on facts, assume the student's interest is represented by the
fuzzy relation P(X,Y):
Interest, P(X,Y) =
Th Ap Hw Pg
Ali 0.2 1 0.8 0.1
Bala 1 0.1 0 0.5
Chin 0.5 0.9 0.5 1
The properties of the courses are indicated by the following fuzzy
relation Q(Y,Z):
Properties, Q(Y,Z) =
FT FC NN ES

www.utm.my
Th 1 0.5 0.6 0.1
Ap 0.2 1 0.8 0.8
Hw 0 0.3 0.7 0
Pg 0.1 0.5 0.8 1
MEEM 1713 – ARTIFICIAL INTELLIGENCE 75
 Using max-min approach:
Thus, P o Q =
FT FC NN ES

AliAli

Bala
Chin

Chin
Ayu

Conclusion,

www.utm.my
1. Ali should choose ........................................
2. Bala should choose ........................................
3. Chin should choose ........................................

MEEM 1713 – ARTIFICIAL INTELLIGENCE 76

 RESULTS OF FUZZY INFERENCE

By using max –min we get:

FT FC NN ES
Ali 0.2 1.0 0.8 0.8
Bala 1.0 0.5 0.6 0.5
Chin 0.5 0.9 0.8 1.0

By using max –product, we get:

FT FC NN ES
Ali 0.2 1.0 0.8 0.8

www.utm.my
Bala 1.0 0.5 0.6 0.5
Chin 0.5 0.9 0.8 1.0

MEEM 1713 – ARTIFICIAL INTELLIGENCE 77

 Exercise 4.7
Suppose we are interested in understanding the speed control
of the DC (direct current) shunt motor under no-load condition.

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 78
 OBJECTIVES
In order to save the armature, you need to find the following:
the minimum and maximum level of Rse ?
the minimum and maximum value of Ia ?
Let Rse be the fuzzy set having a number of possible values of the
series resistance, say n values, given by:

Rse = {Rs1 , Rs 2 , Rs 3 ,..., Rsn }

and let Ia be a fuzzy set having a number of possible values of the

www.utm.my
armature current, say m values, given by:

I a = {I1 , I 2 , I 3 ,..., I m }
MEEM 1713 – ARTIFICIAL INTELLIGENCE 79
 The Fuzzy sets Rse and Ia can be related through a fuzzy
relation, say R, which would allow for the establishment of
various degrees of relationship between pairs of resistance
and current.

In this way, the resistance-current pairings could conform to

the modeler’s intuition about the trade-offs involved in
control of the armature.
 Let W be another fuzzy set having numerous values

www.utm.my
for the motor speed, say v values, given by:

W = {ω1, ω2,…,ωv }

MEEM 1713 – ARTIFICIAL INTELLIGENCE 80

 Using the operation of composition, we could then compute a
relation, say T , to be used to relate series resistance to
motor speed, i.e., Rse to W.

The operations needed to develop these relations are as

follows :- 2 fuzzy cartesian products and 1 composition:
R = R se × I a
S = Ia × W

www.utm.my
T =RοS

MEEM 1713 – ARTIFICIAL INTELLIGENCE 81

 GIVEN
Suppose the membership functions for both series resistance
Rse and current Ia are given in terms of percentages of their
respective rated values, as follows:
µRse (% se ) =
0.3 0.7 1.0 0.2
+ + +
30 60 100 120

µI a (%a ) =
0.2 0.4 0.6 0.8 1.0 0.1
+ + + + +
20 40 60 80 100 120

and the membership value of various motor speeds (W) are:

www.utm.my
µW (rpm ) =
0.33 0.67 1.0 0.15
+ + +
500 1000 1500 1800

MEEM 1713 – ARTIFICIAL INTELLIGENCE 82

 Based on the facts obtained,
develop the relations: R and S?
• First find R:
µI a (%a ) =
0.2 0.4 0.6 0.8 1.0 0.1
+ + + + +
20 40 60 80 100 120

µRse (% se ) =
0.3 0.7 1.0 0.2
+ + +
30 60 100 120
• Min(Rse x Ia)
20 40 60 80 100 120
30 0.2 0.3 0.3 0.3 0.3 0.1

www.utm.my
60 0.2 0.4 0.6 0. 7 0 .7 0.1
R= 
100 0.2 0.4 0 .6 0.8 1 .0 0.1
120 0.2 0 .2 0.2 0.2 0 .2 0.1

MEEM 1713 – ARTIFICIAL INTELLIGENCE 83

 Now S
500 1000 1500 1800
20 0.20 0.20 0.20 0.15
40 0.33 0.40 0.40 0.15
0.33 0.60 0.60 0.15
S= 60  
80 0.33 0.67 0.80 0.15
100  0.33 0.67 1.00 0.15 
120 0.10 0.10 0.10 0.10
For example,

www.utm.my
µ R (60,40 ) = min (0.7,0.4 ) = 0.4, µ R (100,80 ) = min (1.0,0.8) = 0.8,
and µ S (80,1000 ) = min (0.8,0.67 ) = 0.67

MEEM 1713 – ARTIFICIAL INTELLIGENCE 84

 The following relation results from a max-min
composition for T :

500 1000 1500 1800

30  0.30 0.30 030 0.15 
60 0.33 0.67 0.70 0.15
T = Rο S =

www.utm.my
100  0.33 0.67 1.00 0.15 
 
120 0.20 0.20 0.20 0.15

MEEM 1713 – ARTIFICIAL INTELLIGENCE 85

 For instance,

µT (60,1500) = max [min (0.2,0.2),min (0.4,0.4), min (0.6,0.6),

min (0.7,0.8), min (0.7,1.0), min (0.1,0.1)]
= max[0.2,0.4,0.6,0.7,0.7,0.11]=0.7

www.utm.my
Try Max-product next.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 86

 2.5
Prelude to Fuzzy Logic
Control
 Fuzzy Logic
 Linguistic Variables
 Fuzzy Rules

•www.utm.my
 Fuzzy Inference
 Fuzzy Reasoning

MEEM 1713 – ARTIFICIAL INTELLIGENCE

87
 4.5.1 Fuzzy Logic

• The term ‘fuzzy logic’ is used with a number of different meanings.

• Logic means “being reasonable” or “truthful”.
• In one context, it involves the manipulation of fuzzy truth values
such as ‘nearly true’ defined as fuzzy sets over the interval [0,1] of
truth values.
• In fuzzy logic control implementations, natural language statements
can be expressed fairly easily as ‘fuzzy control rules’ which are then
stored in the fuzzy knowledge base of the system.
• The fuzzy rules are normally expressed in a form that will allow the

www.utm.my
rules to be easily programmed (e.g. IF A AND B THEN C).
• Various fuzzy programming tools have been developed to facilitate
fuzzy rule acquisition.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 88

 4.5.2 Linguistic Variables

Many engineering rules can be formulated based on the following

observations and actions:
IF so and so conditions THEN so and so action.

Fuzzy logic allows to a certain extent natural language to be used in

solving many engineering problems as well as other practical
applications.

Since fuzzy logic can handle some form of uncertainty and

www.utm.my
imprecision which are inherent in natural language, thus it can be
used as the mathematical foundation of our natural language.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 89

 Linguistic Hedges and Operators

A fuzzy set can be regarded as corresponding to a linguistic value such as

‘tall’, and a linguistic variable ‘height’ can be regarded as ranging over
such linguistic values.

One powerful aspect of fuzzy sets in this context is the ability to deal with
linguistic quantifiers or ‘hedges’.

Hedges such as “more or less”, “very”, “not very”, “slightly”, etc.,

correspond to modifications in the membership function of the fuzzy set
involved.

The fuzzy set operations such CON, DIL, INT, etc. (see the Table of Hedges

www.utm.my
and Operators) can be used to modify the fuzzy set HIGH such as shown
below.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 90

 Table of Hedges and Operators
.
Hedge Operator definition µ(Hedge) HIGH
Not very
HIGH

Very F CON=F 2
More or Less F DIL=F 0.5
More or less HIGH
HIGH

Plus F F 1.25
Not F 1-F
Not very F 1-CON(F)

www.utm.my
Slightly F INT [Plus F AND Not (Very F)]
A But F A AND F

MEEM 1713 – ARTIFICIAL INTELLIGENCE 91

 Example 4.6

As an example, if x = Speed with U=[0,100], then its term set T(Speed) could
be.
T(Speed) = {Very Slow, Slow, Moderate, Fast,..}.

and the semantic M could be defined as.

M(Fast) = the fuzzy set for “a speed above 90 km/h” with
membership function µFast.

Let U be the universe of the fuzzy variable Speed and fuzzy set Fast on U as.

www.utm.my
 0 0 .1 0 .3 0 .8 1 1 
Fast =  + + + + + 
 5 30 50 70 90 100 

MEEM 1713 – ARTIFICIAL INTELLIGENCE 92

 Then, the membership function of Fast can be represented in the
figure below:

µ SPEED
1
Fast

0 Speed
50 90 110

The other corresponding membership functions for the variable speed can also
be formulated in the following manner:
Very Slow
µ

www.utm.my
SPEED
Slow Average Fast
1

0
20 50 70 90 110 Speed

MEEM 1713 – ARTIFICIAL INTELLIGENCE 93

 4.5.5 Fuzzy Rules
In general, three general forms of fuzzy rules exist for any linguistic
variable. These forms are:
(i) assignment statements.
(ii) conditional statements.
(iii) unconditional statements.
Examples of each of these forms are given as follows:
(i) Assignment Statements.
Temperature is very hot.
(ii) Conditional Statements.
IF x is big THEN y is small.
(iii) Unconditional Statements.
Turn the pressure lower.

www.utm.my
Fuzzy rules are a set of conditional statements as given in case (ii).

Fuzzy control rules are characterized by a collection of fuzzy IF-THEN rules

in which the antecedents and consequents involve linguistic variables.
MEEM 1713 – ARTIFICIAL INTELLIGENCE 94
 Fuzzy Rules
This collection of fuzzy control rules characterizes the simple input-output
relation of the system.

INPUTS OUTPUTS

The general form of the fuzzy control rules in the case of multi-input single
output system (MISO) is:

Ri : IF x is Ai,...., AND y is Bi, THEN z = Ci, i = 1,2,...,n.

www.utm.my
where x,.., y and z are linguistic variables representing the process state
variables and the control variable. Ai,.., Bi and Ci are linguistic values of
linguistic variables x,.., y and z in the universes of discourse U,..V and W,
respectively.
MEEM 1713 – ARTIFICIAL INTELLIGENCE 95
 4.5.5 Fuzzy Inference

Dragon Fly
6 legs
Knowledge
wings
Base
Body with 3 parts
Insect
Infer

www.utm.my
This is a Sensor
dragon fly!

MEEM 1713 – ARTIFICIAL INTELLIGENCE 96

 A fuzzy logic engine

Knowledge
Base

Sensor Inference
SOLUTIONS
readings Engine

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 97
 Example of Intelligent Interpretation

It was unusually Then it must be about

COLD in Japan 0 degrees C, I must
when I was there a wear thick clothings
few days ago. when I go there in this
time of the year.

www.utm.my
Humans can handle linguistic imprecision easily because they can infer
well and have a strong knowledge base.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 98

 A typical fuzzy inference might proceed as follows:

Knowledge: If the water is very hot, then add plenty

of cold water
Fact: The water is moderately hot
Conclusion: Add a little cold water

Example of another fuzzy inference:

Knowledge : If the tomato is red then the tomato is ripe.

www.utm.my
Fact : The tomato is very red.

Conclusion : The tomato is very ripe.

GMP (Generalized Modus Ponens)

MEEM 1713 – ARTIFICIAL INTELLIGENCE 99

 Fuzzy Rule Base

So far we have looked only at single fuzzy rules.

A fuzzy knowledge base usually consists of a number of fuzzy
rules.
No strict formal standard has emerged for the structure of
these fuzzy rules.
In most engineering control applications nowadays, the fuzzy
rules are expressed as ‘IF-THEN’ configuration, e.g. ‘IF x is A
THEN y is B’.

www.utm.my
IF <CONDITIONS> THEN <ACTIONS>.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 100

 There are several sentence connectives, namely, ‘AND’, ‘OR’, and ‘ALSO’, which
are usually allowed.

The connectives ‘AND’ and ‘OR’ are frequently used in the antecedent part of
the fuzzy rules, while the connective ‘ALSO’ is often used in the consequent
part of the fuzzy rules.

In practice, ‘AND’ is usually interpreted as an intersection operator, ‘OR’ is

interpreted as a union operator, and ‘ALSO’ indicates the presence of multiple
outputs in the fuzzy rule.

For instance, consider the kth fuzzy rule in a fuzzy knowledge base expressed

www.utm.my
by:

IF x1 is A k1 OR x 2 is A k2 AND x 3 is A k3
THEN y1 is Bk1 ALSO y 2 is Bk2
MEEM 1713 – ARTIFICIAL INTELLIGENCE 101
 Connectives “AND” and “OR”

Rules can be written in this form:

IF A AND B THEN C.
IF A OR B THEN C.

The AND connective in human language has a direct connection

with the Fuzzy Intersection whereas the OR connective has a
direct connection with Fuzzy Union.

From <Natural Language> to <Mathematical Operations>.

AND  Intersection   or ∧  Min.

www.utm.my
OR  Union  or M
 ∨ Max.
ax.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 102

 IF x1 is A k1 OR x 2 is A k2 AND x 3 is A k3

THEN y1 is Bk1 ALSO y 2 is Bk2

• For simplicity it may be re-written as:

IF A k1 OR A k2 AND A k3 THEN Bk1 ALSO Bk2

• It can be seen that there are two outputs in the consequent part of the
rule.

•www.utm.my
• The fire strength of the rule, which is calculated from the antecedent part,
is applied twice in the computation of y1 and of y2.

MEEM 1713 – ARTIFICIAL INTELLIGENCE

103
 4.5.6 Fuzzy Reasoning
Each rule in the fuzzy knowledge base corresponds to a fuzzy relation.

The fuzzy rule base shown previously calls the relation corresponding to the kth
rule as Rk.

The overall relation, R, of the fuzzy rule base can be obtained by performing a
“union” operation on R1, R2, R3, ……RN.
Consider a MISO system with N rules. Its kth fuzzy rule is given by:

IF A ki AND ... AND A ki AND ... AND A kn THEN Bk

www.utm.my
In general, the fuzzy relation Rk of the kth rule can be expressed as:

Rk = Ak → Bk

MEEM 1713 – ARTIFICIAL INTELLIGENCE 104

 Example of a fuzzy reasoning process

To demonstrate the fuzzy reasoning process using max-min and max-product

compositional operators, let us consider a case where the rule base is given by:
Rule 1 IF x 1 is A 11 AND x 2 is A 21 THEN y is B1
Rule 2 IF x 1 is A 12 OR x 2 is A 22 THEN y is B 2

Two examples are given here: (1) max-min and (2) max-product.

The system antecedents are two inputs x = A’ and the consequent is y = B’.

In fuzzy control engineering, the actual inputs are usually in ‘crisp’ form.

www.utm.my
MEEM 1713 – ARTIFICIAL INTELLIGENCE 105
 Fuzzy reasoning based on max-min operator is usually referred to as
MAX-MIN fuzzy inference due to the minimum operation rule Rc
involved.

While the fuzzy reasoning based on max-product is referred to as MAX-

DOT fuzzy inference due to the product operation rule Rp involved.

The objective of the fuzzy inference procedure is actually to find the

output value such that.:
N
B' =  Bk

www.utm.my
n
k =1
N
•where A' =  A k
k =1
=  A'  R k
k =1

MEEM 1713 – ARTIFICIAL INTELLIGENCE 106

 Rk is a fuzzy relation and Ak is a fuzzy set and thus a compositional
operation has to be used.

Using any one of the inferencing techniques (composition operations),

the output waveforms Bk will be computed.

In many cases several Bk will be obtained where their waveforms are then
combined.

A crisp output value can then be calculated based on the defuzzification

procedure used.

www.utm.my
The defuzzification procedure will be discussed in the next section under
fuzzy logic control.

MEEM 1713 – ARTIFICIAL INTELLIGENCE 107