SOFT COMPUTING (ITE1015)

Module 7: SEARCH STRATEGIES

Dr.B.K.Tripathy
DEAN
School of Information Technology and Engineering
VIT, Vellore-632014
E-mail: tripathybk@vit.ac.in
 SYLLABUS

• Optimization techniques
• Evolutionary Algorithms
• Genetic algorithms
• Traditional Algorithms vs Genetic Algorithms
• Classification of Genetic Algorithms
 OPTIMIZATION- INTRODUCTION

• Optimization techniques are used in every sphere of life now

days
• Knowingly or unknowingly all of us apply it even in our day to
day activities
• When we go to a market to purchase certain items, we have
an objective in our view and a certain number of constraints
attached with them
• We move from store to store if an item satisfying our
constraints and fulfilling the objective are available or not
• In case more number of such items is available we rank them
and go for the best one
 OPTIMIZATION- INTRODUCTION

• The entire industry is following optimisation; maximise

their profit and minimising their cost
• In a sense, optimisation has become a part of all human
endeavours
• The resources which are to be optimized are profit,
quality, time, space, money, infrastructure available etc.
• These resources are always limited
• So, there arise the constraints
• The variables which form a part of the objective are called
decision variables
 OPTIMIZATION- INTRODUCTION
• And the constraints which are imposed on them subject to the
availability, relationship among them are called constraints
• The set of solutions which satisfy the set of constraints form
the solution space
• Mathematically speaking optimization searches the decision
space in pursuit of an array of decision variables that could
produce the maximum acceptability considering the initial
goal
• The classical optimization approaches can be broadly
classified into two categories;
• Direct Method
• Gradient based Methods
 DIRECT METHOD

• Only objective function f(x) and constraint values are used to

guide the search strategy
• As derivative information are not used, these methods are
generally slow, requiring many function evaluations for
convergence
• They can be applied to many problems without much
changes in the algorithm
 MATHEMATICAL FORMULATION OF AN
OPTIMIZATION PROBLEM
• Suppose there are ‘n’ numbers of decision variables
• We denote them together as a vector
• The objective function is denoted by
• This is the case of single objective function
• In fact, there may be more than one objective functions say

• Such a case is called as the multi-objective optimization

problem or multicriteria optimization problem or multi-
attribute optimization problem
• The constraints involving these variables can be of two types;
equality constraints and inequality constraints
 MATHEMATICAL FORMULATION OF AN
OPTIMIZATION PROBLEM
• For definiteness we can assume that all the inequality
constraints are of the form
• If there are constraints of the form then we can transform
them into the form with minor modifications
• We denote the equality constraints as
• the inequality constraints as
• So, the standard form of an optimization problem is:
•
 OPTIMIZATION CONTD…

• It may be noted that the optimisation criteria may be to

maximize an objective function
• But it can be easily transformed to one with minimize easily
• The objective function is also called as the cost function and
the space spanned by the decision variables is called the
solution space or the search space or the response space
• When there are no constraints, i. e. J= K=0, we call the
optimization problem as a unconstrained optimization
problem and the case where only K = 0, we call it an equality-
constrained optimization problem.
• On the other hand, if J=0, we call it as an inequality-
constrained optimization problem.
 OPTIMIZATION CONTD…

• Also, we find that in some optimization methods it is

supposed that it is an inequality based optimization problem
• This is because of the case that we can replace the equality
constraints say by two inequality base constraints which in
turn can be replaced by two constraints
• The objective functions as well as the constraints can be
linear or non-linear functions
• If all of them are linear (That is the highest powers of the
decision variables present in them are 1), we call such an
optimization problem as a linear programming problem
 GRADIENT BASED METHOD
• Use the first or second derivatives of the objective function
and/or constraints to guide the search process
• The critical points of a function f(x) are the solutions of the
′ 𝑑𝑓(𝑥)
𝑓 (𝑥) = 0 or =0
𝑑𝑥
• At a solution point say x = a if 𝑓 ′′ (𝑥) < 0 then ‘a’ is a point of
maximum. If 𝑓 ′′ (𝑥) > 0 then ‘a’ is a point of minimum
• If 𝑓 ′′ (𝑥) = 0 then we go for higher order derivative tests
• Ex: f(x) = Sin(x). Then 𝑓 ′ (𝑥) = cos 𝑥 . So, 𝑓 ′ (𝑥) = 0 when x =
(2n+1)𝜋/2. These are its critical points.
• 𝑓 ′′ (𝑥) = −𝑆𝑖𝑛𝑥. At 𝑥 = 𝜋 2, 𝑓 ′′ (𝑥) = −1 is negative. So, f has
a maximum at this point. At 𝑥 = 3 𝜋 2, 𝑓 ′′ (𝑥) = 1 is positive.
So, f has a minimum at this point.
 SOME OF THE COMMON DIFFICULTIES FACED

• These methods converge quickly to optimal solutions

• But, are not applicable to cases where the objective
functions are
• not differentiable or
• discontinuous problems
• The convergence to an optimal solution depends on the
chosen initial solution
• Most algorithms tend to get stuck to a suboptimal solution
• The solution space may have several optimal solutions, many
of them being locally optimal and generating low objective
function values as the problem variables are mostly
nonlinear and there are complex interactions among these
variables
 SOME OF THE COMMON DIFFICULTIES FACED

• Moreover it is not possible to come out of the local optimal

solutions once the process gets stuck up there

• An algorithm efficient in solving one optimization problem

may not be efficient in solving a different optimization
problem. The traditional optimization algorithms are
developed with a specific type of problem in view

• One algorithm may be best suited for one problem while it

may not even be applicable to a different problem
 SOME OF THE COMMON DIFFICULTIES FACED

• It compels the designers to have knowledge about more than

one optimization methods

• Algorithms cannot be efficiently used on a parallel machine

• Because of the affordability and availability of parallel
computing machines, it is now convenient to use parallel
machines in solving complex engineering design optimization
problems.

• However, since most of the traditional methods use a point-

by-point approach, where one solution gets updated to a
new solution in an iteration, the advantage of parallel
machines cannot be exploited
 SOME OF THE COMMON DIFFICULTIES FACED

• The optimization methods assume that the problem

variables are continuous where as in reality the variables are
discrete only.
• So, in order to make the approaches applicable to these
problems, it is assumed that the variables are continuous
during the process of optimization and at the end a value
closer to the solution is recommended
• The infeasible values of program variables are permitted and
so, the process has to go through all the infeasible solutions,
by the way spending a large amount of time computing them
• This increases the computational time, which is undesirable
 SOME OF THE COMMON DIFFICULTIES FACED

• The nearest upper and lower values for each of the program
variables are to be computed at the end, which leads to the
computation of a total number of such values comes to
• Two options checked for each variable may not guarantee the
forming of the optimal combination with respect to other
variables
• All these difficulties can be eliminated if only feasible values
of the variables are allowed during the optimization process
 BIO INSPIRED ALGORITHMS

• Biologically inspired algorithms (BIAs) mimic the intelligent

behaviours from biologic behaviour of animal, bird, fish and
so on
• Bio inspired algorithms are considered a major part of nature
inspired algorithms
• They have become very important in wide variety of
applications in many domains
• BIAs are motivated by the challenges in applications where
conventional optimization techniques are not effective
 BIO INSPIRED ALGORITHMS

• There are three main branches of BIAs, namely

evolutionary algorithms (EAs), swarm intelligence (SI)
and bacterial foraging algorithms (BFAs), which are
based on the understanding of biological systems,
genetic evolution, animal behaviours and bacterial
foraging patterns
• These algorithms have begun to enjoy great success in
solving complex real-world problems
• In the evolution process, selection of species is made based
on their ability of survival in an environment
 BIO INSPIRED ALGORITHMS

• Reproduction is also found to be accompanied by mutation

of species, which performs as one of the major components
of evolution
• Between the 1950s and 1960s several evolutionary
algorithms have been proposed
• Rochenberg and Schwefel introduced another novel
optimization technique to solve problems in aerospace
engineering, which was named later as evolutionary strategy
• Peter Bienert joined Rochenberg and Schwefel later to
construct an automatic experimenter using simple rules of
mutation and selection. There was no crossover in this
technique
 BIO INSPIRED ALGORITHMS

• Only mutation was used to generate an offspring and an

improved solution was kept at every generation.
• This was essentially a simple trajectory style hill-climbing
algorithm with randomisation.
• In 1966, Fogel et al developed an evolutionary programming
technique by representing the solutions as finite state
machines and randomly mutating one of these machines.
• All the above ideas and innovations have evolved into a much
wider discipline called evolutionary algorithms or
evolutionary computation.
 BIO INSPIRED ALGORITHMS

• The major types of evolutionary algorithms are

• Genetic Algorithm
• Genetic Programming
• Evolutionary programming
• Evolutionary Strategy
 A MOTIVATING EXAMPLE
• When attempting to find a perfect combination of chemicals for
a specific problem, a chemist will undertake a set of experiments
• They know roughly what needs to be achieved but not
necessarily how to achieve it
• A chemist will create a number of experiments, where each
experiment is a combination of different chemicals following
some theoretical basis for the experiments (Search
space/Solution Space)
• The experiments are played out and the promising solutions are
identified and gathered together (Search techniques)
• These new chemical combinations are then used as the basis for
the next round of experiments (Reproduction)
• This procedure is repeated until hopefully a satisfactory
chemical combination is discovered (Iteration)
 GENETIC ALGORITHMS

• Genetic Algorithms are search and optimization procedures

• These are motivated by the principles of natural genetics and
natural selection
• Some fundamental ideas of genetics are borrowed and used
artificially to construct search algorithms that are robust and
require minimal problem information
• Its working principle is very different from that of most of the
classical optimization techniques
 GENETIC ALGORITHMS

• The origin of the theory of evolution is the book

• ON THE ORIGIN OF SPECIES
• written by Charles Darwin

• The central concept lies in the process called

• SURVIVAL OF THE FITTEST
• This principle is used by nature in its selection
 GENETIC ALGORITHMS CONTD…

• All living creatures are descendants of older species

• Any variation occurs is due to natural selection

• some individuals have greater chance of reproduction

• Due to some heritable differences
• And are said to have higher “fitness”
• Fitness measures the success of an organism
 GENETIC MECHANISM

• Genetic mechanism behind evolution : Put forth by Gregory

Mendel in 20th century In biology, a gene is a sequence
• Each cell contains chromosomes of DNA or RNA that codes for a
molecule that has a function

• Are strands (Twisted together) of DNA (Deoxyribonucleic acid)

• Consist of genes (Smallest hereditary unit)
• Genes are inherited form parents
• This (inheritance) happens during reproduction and the
process is called crossover
• The set of all chromosomes is called genome
In humans, genes vary in size from a few hundred DNA bases to more than 2 million
bases.
 GENETIC ALGORITHMS CONTD…

• Both the areas of evolution strategy and the evolutionary

programming technique are still being pursued for research

• The most important aspect which was missed in both these

approaches was recognition of the importance of crossover

• John Holland was the first to propose the concept of

crossover and other related recombination operators

• THESE CONCEPTS FORM THE BASIS OF GENETIC ALGORITHM

 GENETIC ALGORITHMS CONTD…
• These ideas he put forth while working on adaptive systems
in as early as 1962, Real breakthrough in 1975

Edgar Frank
Codd
(1923-2003)
Kalyanmoy Deb David E. Goldberg
John Henry Holland
(1953- )
(1929 -2015)
• Holland published a book in 1975 titled:
• ADAPTATION IN NATURAL AND ARTIFICIAL SYSTEMS
• The contents are based upon his own work and by some of
his colleagues
 REQUIREMENTS FOR EVOLUTION

• Evolution: The theory that all living things have changed in

response to environmental conditions by the natural
selection of randomly occurring mutations

• Mutation: The hypothetical occurrence of new forms arising

through change in gene construction of the nucleus and
differing sufficiently from the parent forms to constitute new
varieties

• FOUR REQUIREMENTS for evolutions to take place

• 1. FITNESS 2.VARIATION 3.REPRODUCTION 4.HEREDITY
 KEY CONCEPTS ELABORATELY

• FITNESS: It measures the ability of an individual to survive

and reproduce
• The difference between the values of one organism to
another is the higher number of offspring of one with the
higher fitness value

• VARIATION: It is essential for evolution to take place

• This is the series of changes occurring in species
• Types of Variation (2):
• Inherited variation
• Environmental variation
 KEY CONCEPTS CONTD…

• INHERITED VARIATION: It is a genetically inherited character

• Off springs inherit half of their characters from each of their
parents.
• EXAMPLE: Colour of hair

• ENVIRONMENTAL VARIATION:
• There are two factors
• Climate
• Culture
• These factors affect organisms
 AN EXAMPLE

• Let us consider a cylindrical can with two design parameters

• h: The height of the can
• d: The diameter of its base
• Let us assume that the can needs to have a volume of at least
300ml
• Objective function of the design is to minimize the cost of the
can material
• The corresponding nonlinear programming problem is
• 𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓(𝑑, ℎ) = 𝑐((𝜋 𝑑2 2) + 𝜋𝑑ℎ)
• The material required = Area of the base +Area of the lid+
Area of the surface
 EXAMPLE CONTD…

• The constraint is 𝑔1 (𝑑, ℎ) = (𝜋𝑑 2 ℎ 4) ≥ 300

𝑑min ≤ 𝑑 ≤ 𝑑max ;
• Variable bounds
ℎmin ≤ ℎ ≤ ℎmax .
• Here ‘c’ is the cost of the can material per square cm.
• The diameter of the can and the height of the can are allowed
to vary in the ranges 𝑑min , 𝑑max and ℎmin , ℎmax
respectively
 SOLUTION REPRESENTATION

• The first step in finding a solution to the above NLP is to

represent the parameter values in binary strings
• Suppose we use 5 bit representation of the design
parameters d and h
• Overall string length is 10
• For example if the diameter is 8 cms and height is 10 cms
• Then the string representation of the can is 01000 01010
• The diagrammatic representation of this can is
• (d, h) = (8, 10) Area
ℎmin , ℎmax ℎmin , ℎmax ℎmin , ℎmax
• Chromosome = 𝝅. 𝟖 𝟐 𝟐] + 𝝅. 𝟖. 𝟏𝟎
23 =
• 01000 01010 ℎ , ℎ ℎ ,ℎ ℎ ,ℎ
= 𝟏𝟕𝟔𝒔𝒒 𝒄𝒎
min max min max min maxCost= c. 176 = 23
C = 0.0654
 EXAMPLE CONTD…

• The lower and upper bounds of both d and h are taken to be

0 and 31 respectively. The lower and upper bounds allow GAs
to consider only integer values in the range [0, 31]

• Coding the parameters in a binary string is primarily used in

order to have a pseudo-chromosomal representation of a
design solution

• The 10-bit string illustrated above can be considered to show

a biological representation of a can having 8 cm diameter and
10 cm height
 EXAMPLE CONTD…

• There are two representations for an entity

• phenotypic representation
• genotypic representation
• Looking at a person provides the phenotypic representation
• Each of the features of the person is precisely written in
his/her chromosome - the genotypic representation
• For the can the phenotypic representation is 32
• The genotypic representation is 01000 01010 (artificial
chromosome
 EXAMPLE

• In the can design problem,

• The can itself is the phenotypic representation of an artificial
chromosome of 10 genes
• To see how these 10 genes control the phenotype (the shape)
of the can, let us investigate the leftmost bit (gene) of the
diameter (d) parameter
• A value of 0 at this bit (the most significant bit) allows the can
to have diameter values in the range [0, 15] cm, whereas
• The other value 1 allows the can to have diameter values in
the range [16, 31] cm
• This bit (or gene) is responsible for dictating the slimness of
the can
 EXAMPLE

• Each bit position or combination of two or more bit

positions can also be explained to have some feature of the
can, but some are interesting and important and some are not
that important
• we are ready to apply some genetic operations to such
strings to hopefully find better and better solutions
• Before we do that, we shall describe another important step
of assigning a 'goodness' measure to each solution
represented by a string
 ASSIGNING FITNESS TO A SOLUTION
• GAs work with strings representing design operators instead
of parameters themselves
• Once a string is created by a genetic operator, it is necessary
to evaluate the solution
• This is done under the context of the objective function and
constraint functions
• In the absence of constraints, the fitness of string is assigned a
value which is a function of the solution’s objective function
value
• In most of the cases it is the objective function value
• In the example of the can (taking c = 0.0654)
• 𝐹(𝑠) = 0,654(𝜋 8 2 + 𝜋(8)(10)) = 23
• Keeping the objective in view, smaller fitness value is better
 GENETIC OPERATORS

• We now discuss the genetic operators, in the light of the can

design problem
• Let us consider phenotypes of a random population of six
cans as below
 GENETIC OPERATORS
• There are 3 genetic operators:
 Reproduction Operator
 Crossover Operator
 Mutation Operator
 REPRODUCTION OPERATOR
• The primary objective of the reproduction operator is to
emphasize good solutions and eliminate bad solutions in a
population, while keeping the population size constant

• Identifying good (usually above-avg.) solutions in a

population

• Making multiple copies of good solutions

• Eliminating bad solutions from the population so that
multiple copies of good solutions can be placed in the
population
 REPRODUCTION OPERATOR- METHODS

• There are several methods to achieve the above

 Tournament selection
 Proportionate selection
 Ranking selection

• Binary tournament selection

• Played between two solutions
• The better solution is chosen and placed in a population slot
• Two other solutions are picked again
• Another population slot is filled up with the better solution
 BINARY TOURNAMENT SELECTION CONTD…

• If done systematically, each solution can be made to

participate in exactly two tournaments
• The best solution in a population will win both times, thereby
making two copies of it in the new population
• The worst solution will lose in both tournaments and will be
eliminated from the population
• From a given population, new population will have zero, one,
or two copies of solutions
• Tournament selection has better convergence and
computational time complexity properties compared to any
other reproduction operator that exists in the literature
BINARY TOURNAMENT SELECTION APPLIED TO
EXAMPLE
POPULATION AFTER REPRODUCTION
 CROSSOVER OPERATOR

• Reproduction operator cannot generate new solutions

• It only makes more copies of good solutions at the expense
of not-so-good solutions
• Creation of new solutions is performed in crossover and
mutation operators
• There exist a number of crossover operators in the GA
literature
• Single point crossover, Two point crossover
• Multipoint crossover, Uniform crossover etc
• In almost all crossover operators, two strings are picked from
the mating pool at random and some portions of the strings
are exchanged between the strings
 CROSSOVER CONTINIUED

• In the example applying single point crossover, we get

• It is true that every crossover between any two solutions from

the new population is not likely to find children solutions
better than parent solutions, but the chance of creating
better solutions is far better than random
• If bad solutions are created, they get eliminated in the next
reproduction operator and hence have a short life
 CROSSOVER RESTRICTIONS

• In order to preserve some good strings selected during the

reproduction operator, not all strings in the population are
used in crossover
• If a crossover probability of p is used then 100p % strings in
the population are used in the crossover operation and
100(1 -p)% of the population are simply copied to the new
population
SINGLE POINT CROSS OVER
 TWO POINT CROSS OVER

• Here two random sites are chosen

• The contents bracketed by these sites are exchanged between
two mated parents
• If the cross-site 1 is six and cross-site 2 is twelve, the strings
between six and twelve are exchanged
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Parents
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

0 0 0 0 1 1 1 1 1 1 1 0 0 0 0 0 0
0
Children
1 1 1 1 1 0 0 0 0 0 0 0 1 1 1 1 1 1
 MULTI POINT CROSS OVER

• In a multi-point cross over, again there are two cases.

• Even number of cross-sites
• Odd number of cross-sites
• In case of even numbered cross-sites, the string is treated as
a ring with no beginning or end
• The cross-sites are selected around the circle uniformly at
random
Parent 1 Parent 2 Child 1 Child 2

Before Crossing
After Crossing
 MULTI POINT CROSS OVER CONTD…

• Odd number of Cross sites

Parent 1
Parent 2

Child 1
Child 2
 UNIFORM CROSS OVER

• An extreme of multi-point cross over

• Without Mask:
• Each bit from either parent is selected with a probability of
0.5 and then interchanged
• It is radically different from one-point cross over
• With Mask:
• Sometimes gene in the offspring is created by copying the
corresponding gene from one or the other parent chosen
according to a randomly generated cross over mask
 UNIFORM MASK CONTD…

• Mask comprises of 1s and 0s

• Mask bit = 1: gene is copied from the first parent
• Mask bit = 0: gene is copied from second parent
• To produce the second offspring: The parents are exchanged
• A NEW CROSS OVER MASK IS GENERATED FOR EACH PAIR OF
PARENTS
• Offspring therefore contains a mixture of genes from each
parent
• The number of effective crossing points is not fixed but
averages to L/2 (where L is the chromosome length)
 UNIFORM CROSSOVER CONTD…
(1, 0) pair
Parent 1 1 0 0 1 1 0 0 0 1 0 1 0 1 1 0 0 1 1
becomes (0,
1). Also,
Parent 2 0 0 0 1 0 0 1 1 0 0 1 1 0 0 1 0 0 1
conversely.
Child 1 0 0 0 1 0 0 1 1 0 0 1 1 0 0 1 0 0 1 (0, 0) or (1,
1) remains
Child 2 1 0 0 1 1 0 0 0 1 0 1 0 1 1 0 0 1 1 unchanged
Uniform Crossover without mask

With Mask 1 0 0 1 0 1 1 1 0 0 1 0 1 0 0 1 0 0
Parent 1 1 0 0 1 1 0 0 0 1 0 1 0 1 1 0 0 1 1

Parent 2 0 0 0 1 0 0 1 1 0 0 1 1 0 0 1 0 0 1

1 0 0 1 0 0 0 0 0 0 1 1 1 0 1 0 0 1
Child 1
Child 2 0 0 0 1 0 0 0 0 0 0 1 1 1 0 1 0 0 1

If in the mask the bit is 1, bit from the first parent is selected. Else, from parent 2 to
generate the first child. For the second child the roles of parents are interchanged
 MUTATION

• The mutation operator changes a 1 to a 0 and vice versa with

a small mutation probability, Pm.
• The need for mutation is to keep diversity in the population

• the solution obtained is better than the original solution.

Although, it may not happen all the times, mutating a string
with a small probability is not a random operation since the
process has a bias for creating a few solutions in the
neighbourhood of the original solution
 THE BASIC GENETIC ALGORITHM

• Step 1: Start with a randomly generated population

• Step 2: Calculate the fitness of each chromosome in the
population
• Step 3: Repeat the following steps until n offsprings have
been created:
3.1 Select a pair of parent chromosomes from the current
population
3.2 With probability 𝑃𝐶 crossover the pair at a randomly
chosen point to form two offsprings
3.3 Mutate the two offsprings at each locus with
probability 𝑃𝑚
• Step 4: Replace the current population with the new
population
• Step 5: Go to Step 2
 THE FLOW CHART OF GA
g

BEGIN

INITIALIZE POPULATION
t=0

t=t+1 COND?

REPRODUCTION

CROSSOVER

MUTATION
 CONVERGENCE CONDITION

• A genetic algorithm is usually said to converge when there is

no significant improvement in the values of fitness of the
population from one generation to the next
 GENETIC MECHANISM CONTD…

• Each gene encodes a different trait (distinguishing

characteristic)
• There is a different variety of gene called the allele
• For Example:
• There are different alleles for brown hair and dark hair

• The flow of gene in a population or

• Transfer of alleles or genes from other populations or due to
mutation,
• which is a change in the structure of a gene that is usually the
consequence of an error in the transcription of DNA
 GENETIC ALGORITHMS

• Developed during late 1950s and early 1960s

• By some evolutionary biologists
• As a result of their trials to develop algorithms to model
aspects of natural evolution through programming in
computers
• Evolutionary computing inherent
• Biologists did not realise then that their strategy would be
applicable to artificial problems
• Evolution- inspired algorithms were independently
developed by
• G.E.P. Box, G.J. Friedman, W.W. Bledsoe and H.J.
Bremermann
 GENETIC ALGORITHMS

• These algorithms were applied by them for function

optimization and machine learning
• This was around 1962 and was not pursued further
immediately
• The next step in the development was by I. Rechenberg in
1965, when he introduced a technique called EVOLUTION
STRATEGY
• It had very little similarity to genetic algorithms
• Most of the concepts used in genetic algorithms like
population, crossover and mutation of parents to produce
offspring were not there
 GENETIC ALGORITHMS

• The technique of evolutionary programming was introduced

by three scientists in 1966
• L.J. Fogel, A.J. Owens and M.J. Walsh

• It is the next important step in this direction

• The solutions were represented by finite state machines
• It follows the evolution strategy of Rechenberg
• The process was:
• To randomly mutate one of these simulated machines and
keeping the better of the two
 GENETIC ALGORITHMS CONTD…

• Both the areas of evolution strategy and the evolutionary

programming technique are still being pursued for research

• The most important aspect which was missed in both these

approaches was recognition of the importance of crossover

• John Holland was the first to propose the concept of

crossover and other related recombination operators

• THESE CONCEPTS FORM THE BASIS OF GENETIC ALGORITHM

 GENETIC ALGORITHMS CONTD…

• These ideas he put forth while working on adaptive systems

in as early as 1962
• Real breakthrough in 1975

• He published a book titled:

• ADAPTATION IN NATURAL AND ARTIFICIAL SYSTEMS

• The contents are based upon his own work and by some of
his colleagues
 GENETIC ALGORITHMS CONTD…

• This book encompasses the idea of adaptive digital systems

and the three ingredients of genetic algorithm

• Crossover
• Mutation
• Selection

• Defined and used

• Thus the problem solving strategy using biological evolution
was simulated
 GENETIC ALGORITHMS CONTD…

• These three operators are simple and straightforward

• The reproduction operator selects good strings
• The crossover operator recombines good substrings from
two good strings together to form a hopefully better substring
• The mutation operator alters a string locally to create a
better string
• Even though none of these claims are guaranteed and/or
tested during a GA generation, it is expected that if bad
strings are created they will be eliminated by the
reproduction operator in the next generation and if good
strings are created, they will be emphasized.
 GENETIC ALGORITHMS CONTD…

• The theoretical soundness of genetic algorithm could be

established through the introduction of the notion of
schemata.
• Further, the dissertation of K. D. Jong carried the ideas
forward by showing that GA can perform well on variations
like over a wide range of test functions, and noisy and
discontinuous searches

• Several published books; like Davis (1991), Goldberg (1989),

Holland (1975), Michalewicz (1992) and Deb (1995) contain
most of the GA studies available till the corresponding time.
 GENETIC ALGORITHMS CONTD…

• Several conference proceedings update the knowledge and

contributions at regular intervals
• Many efforts have been made to apply GA to structural
engineering problems
• This includes works of Goldberg and Samtani (1986), Jenkins
(1991), Deb (1991) and Rajeev and Krishnamurthy (1992)
• But the applications of GA are not limited to structural
engineering only
• Several applications have been obtained in the fields of
biology, computer science, image processing and pattern
recognition, physical science, social sciences and neural
networks
 GENETIC ALGORITHMS CONTD…

• At first, these applications were mainly theoretical

• As research continued to proliferate, genetic algorithms

migrated into the commercial sector

• Their rise fueled by the exponential growth of computing

power and the development of the Internet

• Today, evolutionary computation is a thriving field

 GENETIC ALGORITHMS CONTD…

• Genetic algorithms are "solving problems of everyday

interest"

• in diverse areas as
• stock market prediction and portfolio planning
• aerospace engineering
• microchip design
• biochemistry and molecular biology
• scheduling at airports and assembly lines
 GENETIC ALGORITHMS CONTD…

• The power of evolution has touched virtually any field one

cares to name
• Shaping the world around us invisibly in countless ways
• New uses continue to be discovered as research is ongoing

• And at the heart of it all lies nothing more than Charles

Darwin's simple, powerful insight:
• That the random chance of variation, coupled with the law
of selection, is a problem-solving technique of immense
power and nearly unlimited application
 GENETIC ALGORITHMS CONTD…

• The language used to describe the elements of a genetic

algorithm is directly derived from biology, from which they
draw inspiration
• The set of all possible solutions to the problem is called the
search space, or population
• Each feasible solution is an individual

• Each feasible solution is characterised by two factors:

• A chromosome, which is typically encoded as a bit string
• A value of fitness
 GENETIC ALGORITHMS CONTD…

• Short blocks of different bits of a chromosome that encode a

different parameter of the solution are the genes, which can
have different values, or alleles: in a bit string an allele is 0 or
1
• The position of a gene in a chromosome is the locus
• The collection of all the genes is the genotype
 GENETIC ALGORITHMS CONTD…

• In a genetic algorithm
• Some individuals are chosen as the parents
• Selection is based on their fitness
• Individuals with higher fitness are more likely to be selected
• The parents are then allowed to produce offspring
• By applying reproductive operators
• Reproductive operators are
• Crossover
• Mutation
• Crossover operators recombine the chromosomes of the
parents
• The most common type is called single point crossover
 GENETIC ALGORITHMS CONTD…

• A random point in which to cut the two genes is selected

• The first part of the first parent is then combined with the
second part of the second parent to create a child

• Similarly, the second part of the first parent is combined with

the first part of the second parent to produce the second
child
 TWO IMPORTANT CONCEPTS

• Chromosome: A microscopic, thread like bundle of Deoxyribo

Neucleic Acid (DNA) molecules which collectively carry the
hereditary material in subunits called genes

• Mutation: The process by which new forms arise

By a changing or being changed the hypothetical
occurrence of new forms
Arising through change in gene construction of the
nucleus and
Differing sufficiently from the parent forms to
constitute new varieties
 GENETIC ALGORITHMS CONTD…

• Crossover does not always occur

• Its probability is just between 60% and 70%

• Mutation is rarer
• It occurs with a probability between 0.01% and 0.02%
• When it happens, one bit changes from 0 to 1 or vice
versa
• Mutation assures that the population doesn’t become
too static and therefore incapable of reproduction
 GENETIC ALGORITHMS CONTD…

• The off springs form a new generation and the process starts
from the beginning
• Again, fitness is evaluated for each individual of this new
population
• The selected ones reproduce until a certain number of new
generations have been produced and analyzed
• The problem is then satisfied
 GENETIC ALGORITHMS CONTD…

• Genetic algorithms are

• Good at taking larger, potentially huge search spaces and
• navigating them looking for optimal combinations of things
and
• solutions which we might not find in a life time
• Genetic algorithms are very different from most of the
traditional optimization methods
 GENETIC ALGORITHMS CONTD…

• Genetic algorithms need design space to be covered into

genetic space
• So, genetic algorithms work with a coding of variables
• The advantage of working with a coding of variable space is
that
• Coding discretizes the search space even though the function
may be continuous
 GENETIC ALGORITHMS

• A more striking difference between genetic algorithms and

most of the traditional optimization methods is that
• GA uses a population of points at one time in contrast to the
single point approach by traditional optimization methods
• This means that GA processes a number of designs at the
same time
• To improve the search direction in traditional optimization
methods, transition rules are used and they are deterministic
in nature but GA uses randomized operators
• Random operators improve the search space in an adaptive
manner
 GENETIC ALGORITHMS

• Three most important aspects of GA are

 Definition of objective function
 Definition and implementation of genetic representation
 Definition and implementation of genetic operators

• Once these three have been defined, the GA should work

fairly well beyond doubt
• We can, by different variations, improve the performance,
find multiple optima (species if they exist) or parallelize the
algorithms
 REPRODUCTION IN GA

• Is usually the first operator applied on population

• Chromosomes are selected from the population to be parents
to cross over and produce offspring
• According to Darwin’s evolution theory the best ones should
survive and create new offspring
• That is why reproduction operator is sometimes known as the
SELECTION OPERATOR
• There are a number of reproduction operators in GA
• The essential idea is that the above average strings are picked
from the current population and
• Their multiple copies are inserted in the mating pool in a
probabilistic manner
 SELECTIONS IN GA

• The various methods of selecting chromosomes for parents to

cross over are

• Roulette-wheel selection
• Boltzmann selection
• Tournament selection
• Rank selection
• Steady-state selection
 ROULETTE-WHEEL SELECTION

• The commonly used reproduction operator is the

proportionate reproductive operator where a string is
selected from the mating pool with a probability
proportional to the fitness
• Thus, i string in the population is selected with a probability
th

proportional to Fi where Fi is the fitness value for that

string
• Since the population size is usually kept fixed in a simple GA,
the sum of the probabilities of each string being selected for
the mating pool must be one
F
• The probability of the i selected string is
th p
i  n
i

F
j 1
j
 ROULETTE-WHEEL SELECTION CONTD…

• One way to implement this selection scheme is to imagine a

Roulette-wheel with its circumference for each string marked
proportionate to string’s fitness

• The fitness of the population is calculated as Roulette-wheel

is spun ‘n’ times (usually n = 8), each time selecting an
instance of the string chosen by the Roulette-wheel pointer.

• Since the circumference of the wheel is marked according to

a string’s fitness, the Roulette-wheel mechanism is expected
to make Fi / F copies of the ith string of the mating pool.
A ROULETTE-WHEEL
 ROULETTE-WHEEL SELECTION CONTD…

• The average fitness 1 n 

F    Fj 
n  j 1 

This Roulette-wheel selection scheme can be simulated easily.

• Using the fitness value Fi of all strings, the probability of
selecting a string being copied, can be calculated by adding
the individual probabilities from the top of the list.
• The Roulette-wheel concept can be simulated by realizing that
the string in the population represents the cumulative
probability.
• Thus the first string represents the cumulative values from 0 .
• Hence cumulative probability of any string lies between 0 – 1.
 ROULETTE-WHEEL SELECTION CONTD…

• In order to choose n strings, n random numbers between

zero and one are created at random.
• Thus, the string that represents the chosen random number in
the cumulative probability range (calculated from fitness
value) for the string, is copied to the matting pool.
• This way, the string with a higher fitness value will represent a
larger range in the cumulative probability values and
therefore, has a higher probability of being copied into the
mating pool.
• On the other hand, a string with a smaller fitness value
represents a smaller range in cumulative probability values
and has a smaller probability of being copied into the mating
pool.
 BOLTZMANN SELECTION

• Simulated annealing is a method of functional minimization

or maximization.
• This method simulates the process of slow cooling of molten
metal to achieve the minimum function value in a
minimization problem.
• The cooling phenomenon is simulated by controlling a
temperature like parameter introduced with the concept of
Boltzmann probability distribution
• A system in thermal equilibrium at a temperature T has its
energy distributed probabilistically according to
P( E )  exp( E / kT )
where ‘k’ is Boltzmann constant
 GENETIC ALGORITHMS

• This expression suggests that a system at a

• High temperature has almost uniform probability of being at
any energy state, but at a
• Low temperature it has a small probability of being at a high
energy state.

• Therefore, by controlling the temperature T and assuming

search process follows Boltzmann probability distribution, the
convergence of the algorithm is controlled.

• THIS IS BEYOND OUR SCOPE AND DETAILS CAN BE FOUND IN

THE BOOK BY DEB IN 1995.
 TOURNAMENT SELECTION

• GA uses a strategy to select the individuals from population

and insert them into a mating pool.
• Individuals from the mating pool are used to generate new
offspring, which are the basis for the next generation.
• As the individuals in the mating pool are the ones whose
genes will be inherited by the next generation, it is desirable
that the mating pool consists of good individuals.
• A selection strategy in GA is simply a process that favours
the selection of better individuals in the population for the
mating pool.
• There are two important issues in the evolution process of
genetic search, population diversity and selective pressure,
as given by Whitley in 1989.
 GENETIC ALGORITHMS

• Population diversity means that the genes from the already

discovered good individuals are exploited while promising the
new areas of the search space continue to be explored.

• Selective pressure is the degree to which the better

individuals are favoured.

• The higher the selective pressure the more, the better

individuals are favoured.

• THE SELECTIVE PRESSURE DRIVES GA TO IMPROVE

POPULATION FITNESS OVER SUCCEEDING GENERATIONS.
•
 GENETIC ALGORITHMS

• The convergence rate of GA is largely determined by the

selective pressure and population diversity.

• In general, higher selective pressure results in higher

convergence rates.

• However, if the selective pressure is too high, there is an

increased chance of GA prematurely converging to local
optimal solution because the population diversity of the
search space to be exploited is lost.
 RANK SELECTION
• The Roulette-wheel will have problem when the fitness
values differ very much
• For example, if the best chromosome fitness is 90%, its
circumference occupies 90% of Roulette-wheel, and then
other chromosomes will have very few chances to be
selected.
• Rank selection first ranks the population and taken every
chromosome, receives fitness from the ranking.
• The worst will have fitness 1, the next 2, ..., and the best will
have fitness N (N is the number of chromosomes in the
population).
• The Roulette-wheel selection is applied to the modified wheel
 STEADY- STATE SELECTION

• This is not a particular method of selecting the parents

• The main idea of the selection is that
• BIGGER PART OF CHROMOSOME SHOULD SURVIVE TO THE
NEXT GENERATION

• Here, GA works in the following way

• In every generation are selected, a few (good individuals with
high fitness for maximization problem) chromosomes, for
creating new off springs
• Then, some (bad with low fitness) chromosomes are
removed and new offspring is placed in that place

• THE REST OF THE POPULATION SURVIVES IN A NEW

GENERATION
 CROSSOVER AND MUTATION PROPERTIES

• After the reproduction phase is over, the population is

enriched with better individuals

• Reproduction makes clones of good strings, but does not

create new ones

• Cross over operator is applied to the mating pool with a

hope that it would create a better string

• The aim of the cross over operator is to search the parameter

space
• In addition, search is to be made in a way that the information
stored in the present string is maximally preserved because
these parent strings are instances of good strings selected
during reproduction
 CROSSOVER AND MUTATION PROPERTIES

• Cross over is a recombination operator, which proceeds in

three steps
 The reproduction operator selects at random a pair of
individual strings for mating
 A cross site is selected at random along the string length
 The position values are swapped between two strings
following the cross site
• Cross over rate is the probability of cross over and it is
denoted by

• This probability is normally fixed as 0.5 to 1 for population size

of 30 to 200.
 CROSS OVER RATE

• Accordingly the number of pairs to be crossed is decided

• We have seen that with random cross-sites, the children
strings produced may not have a combination of good
substrings from parent strings depending on whether or not
the crossing site falls in the appropriate place
• But we do not worry about this too much because if good
strings are created by cross over, there will be more copies of
them in the next mating pool generated by the reproduction
operator
• But if good strings are not created by cross over, they will
not survive too long, because reproduction will select
against those strings in subsequent generations
 CROSS OVER RATE

• It is clear from this discussion that the effect of cross over

may either be detrimental or beneficial
• Thus, in order to preserve some of good strings that are
already in the mating pool, not all strings in the mating pool
are used in cross over
• When a cross over probability of PC is used only 100 percent
strings in the population are used in the cross over operation
and 100(1- PC) percentage of the population remains as it is in
the current population
• Even though the best 100(1- PC )% of the current population
can be copied deterministically to the new population, this is
usually preferred at random
• A CROSS OVER OPERATION IS MAINLY RESPONSIBLE FOR THE
SEARCH OF NEW STRINGS
 FITNESS FUNCTION

• As pointed earlier, GA mimics the Darwinian theory of

survival of the fittest and principle of nature to make a search
process.

• Therefore, GAs are usually suitable for solving maximization

problems

• Minimization problems are usually transformed into

maximization problems by some suitable transformation
• In general, fitness function F(X) is the first derived from the
objective function and used in successive genetic operations.
 FITNESS FUNCTION

• Certain genetic operations require that fitness function be

non-negative, although certain operators do not have this
requirement
• Let f(X) be the objective function
• Consider the following transformations:
• F(X) = f(X) for maximization problem
• F(X) = 1/f(X) for minimization problem, if f(X)  0
• F(X) = 1/(1+f(X)), if f(X) = 0
• A number of such transformations are possible
• The fitness function value of the string is known as string’s
fitness
 APPLICATIONS OF GENETIC ALGORITHMS

• Genetic algorithms can be applied in a variety of different

situations in science, engineering, but also in economics and
social sciences.

• Combinatorial optimization problems: an example of this is

the Travelling Salesman Problem
• The aim, given a list of cities and their distances, is to find the
shortest route to visit each city once and then return to the
starting point
• The benefits of the use of genetic algorithms are linked to the
mutation and crossover operators
• The former allows a parallel search, while the latter allows
individuals to share information
 APPLICATIONS OF GENETIC ALGORITHMS

• Economics: genetic algorithms are applied in Game theory

(During the last 17 years 13 peoples working in game theory
have received Nobel prize in Economics)
• A field of economics that studies strategic decision making
• The cooperation and conflicts between rational agents making
decisions

• They are also applied in the game of prisoner’s dilemma, a

game that illustrates the possibility that two rational beings
might decide not to cooperate, even if it seems their best
interest to do so
 PRISONER’S DILEMMA

• The Prisoner's Dilemma is a classic problem in game theory.

• It has the paradoxical outcome that members of a group will
consciously steer towards a sub-optimal outcome in certain
scenarios
• The game is usually phrased in terms of two suspects, both of
whom have been arrested for a major crime, who are offered
a bargain.
• If both stay silent, each of them can still be convicted of a
minor crime and sentenced to 6 months in prison.
 PRISONER’S DILEMMA

• If one of them confesses and implicates the other, this

provides evidence of a major crime.

• The confessor is rewarded by being let off of all crimes, and

the other suspect will serve ten years in prison.

• If both confess, they will both serve two years in a plea for the
major crime

• It is obvious that the best outcome (the Pareto optimum) for

the group would be if both prisoners cooperated and stayed
silent: Six months for both prisoners.
 PRISONER’S DILEMMA

• However, in the "default" setting of the Prisoner's

dilemma,
• we assume that the prisoners are not given the chance
to work out such a strategy and that they are interested
in their own wellbeing first.

• Prisoner A will now analyse his options:

• If Prisoner B chooses "don't confess", Prisoner A's best
choice will be "confess": A gets out of prison
immediately.
• If Prisoner B chooses "confess", Prisoner A's best choice
will be "confess", too: 2 years is better than 10 years.
 PRISONER’S DILEMMA

• Using this reasoning, both prisoners will choose "confess" as

providing the best outcome for themselves in all
circumstances, even though it is not best result for the group.

• The strategy "confess" is a strictly dominant strategy: The

choice of the Prisoner B does not change the way Prisoner A
will act. The "confess/confess" scenario is also the only Nash
equilibrium in this problem.

• When a situation like the prisoner's dilemma, where the

Pareto efficient outcome is not the actual outcome, occurs in
a market economy, that can be an example of a market failure
 APPLICATIONS IN ECONOMICS

• Genetic algorithms can be used to explore the possibility of

developing cooperative behaviour
• Economics uses the following features of genetic algorithms:
the ability to represent individuals as part of a population
whose agents differ in strategy, the processing of information
in parallel and the selection of better-performing strategies
 APPLICATIONS OF GENETIC ALGORITHMS

• Scheduling: GAs are often applied in order to solve

production scheduling problems because they operate in a set
of various solutions in parallel

• The population of solutions consists in different and often

conflicting answers

• The goal of the genetic algorithm is to find the best way to

assign the time to the activities to maximize the overall
performance
 APPLICATIONS OF GENETIC ALGORITHMS

• Robotics: genetic algorithms are applied in this field to obtain

good navigating systems

• The navigation system must be able to direct the robot in a

constricted environment and get it to its destination, without
it crashing into objects and getting lost

• The genetic algorithm scans the intermediary points of every

candidate path to find the most direct move to the
destination; when an obstacle is found, it stops and goes
back to the last free path section analysed.
 APPLICATIONS OF GENETIC ALGORITHMS

• It is proven that genetic algorithms are able to perform very

well in real life and complex situations, because they are
adaptive to the environment to which they are applied, just
like real life organisms adapt to the natural environment in
which they find themselves.
• The study of GAs is also used to model
• Ecosystems
• Immune systems
• Cognitive systems
• social systems
 APPLICATIONS OF GENETIC ALGORITHMS

• Industrial design by parameterization

• Network design by construction

• Routing – finding the shortest path to deliver the packets,

perform dynamic and anticipatory routing in
telecommunications, easy of switching and optimize the cell
tower placement with best coverage of signals

• Time series prediction – learning the strategies based on the

history of data
• Database mining – classification, prediction, association
• Control systems
• Artificial life systems
 APPLICATIONS OF GENETIC ALGORITHMS

• Chemistry : molecular conformation

• Automotive design – GA is used to design the four wheeler
vehicles that are efficient in shape, fuel consumption and
safety

• The modelling done by GA provides more options for the

designers in less cost

• Engineering design – optimize the structural and operational

designs in machineries, factories, buildings for optimization
of heat exchangers, robot gripping arms, satellite booms,
building trusses, turbines etc., It is also used to identify the
fault
 APPLICATIONS OF GENETIC ALGORITHMS

• Evolvable - design of circuits in VLSI design, embedded

systems, reconfiguring circuits which are stochastic to evolve
new configuration from the old model

• Routing – selecting the most efficient and cost effective

routing for shipment, air traffic, trip plan packet by travel
agency, improving the productivity etc.

• Gaming – GA are programmed to find the most successful

strategies from the previous game histories. It learns from its
experience in gaming that uses game theory
• Security – encryption for sensitive data and breaking those
codes. Using GA one build or break the complex code for
security
 APPLICATIONS OF GENETIC ALGORITHMS

• Molecular design – used in applied chemistry and

medicine for understanding of molecular structure,
protein functions, predicting the binding between the
protein molecules used by pharmaceutical industry to
invent medicine for particular diseases

• Gene Expression - analysis of gene expression in micro

array data helps in classifying the genes or to identify the
genetic cause for the disease which moved treating
patients based on personalized medicine
 APPLICATIONS OF GENETIC ALGORITHMS

• Finance Management – market trend prediction, study of

economic meltdown, assisting in investments by
understanding the strategies using GA

• Marketing – GA helps the merchandisers in production and

marketing consultants in planning and advertisement to sell
the products
HYBRID SYSTEMS
 WHY HYBRID SYSTEMS?

• Every system has some positive features and some negative

features
• So, when we can combine two or more basic systems by
taking most of their positive features then we are supposed to
get a hybrid system which outperforms the individual systems
• An Example:
• When rough set was introduced in 1982, it was supposed by
many to be a competitive model to fuzzy sets
• In a paper entitled “Fuzzy sets vs Rough sets” Pawlak tried to
provide a comparison of the two models of uncertainty and
show that rough set is a better model than fuzzy sets
 EXAMPLE CONTD…

• However, it was shown by Dubois and Prade in 1990 in their

paper, ”Rough Fuzzy sets and Fuzzy rough Sets”, International
Journal of General Systems that Fuzzy set and rough sets are
not competing models.
• These are complementary models
• They combined these two models to propose Fuzzy rough
sets and Rough fuzzy sets
• It has been observed that in real life applications these two
hybrid models are better than the individual models as is
seen in the development of rough fuzzy c-means clustering
algorithms
 FORMATION OF RULES

• The general way of representing human knowledge is by forming

natural language expressions given by
• IF antecedent THEN consequent
• There are 3 general forms that exist for any linguistic variable
• 1. Assignment statements [ limit the value of a variable to a specific
quantity]
• 2. Conditional statements
• 3. Unconditional statements
• 1. Examples:
• Y = small
• Orange colour = orange
• Climate = Autumn
 FORMATION OF RULES CONTD…

2. Conditional statements
IF y is very cool THEN stop
IF A is high THEN B is low ELSE B is not low
IF temperature is high THEN climate is hot
3. Unconditional statements
Goto A
STOP
Turn the pressure low
 FORMATION OF RULES CONTD…

• Canonical form of fuzzy rule base systems:

• Rule 1: If conclusion C1 then restriction R1
• Rule 2: If conclusion C2 then restriction R2
• ……
• ……
• Rule n: If conclusion Cn then restriction Rn
• Generally both unconditional as well as conditional
statements place some restrictions on the consequent of the
rule base process
• Fuzzy sets and relations generally model these restrictions
 FORMATION OF RULES CONTD…

• The restriction statements irrespective of conditional or

unconditional statements are usually connected by logical
connectives such as “and”, “or” or “else”
• The restrictions denoted by R1, R2,… Rn apply to the
consequent of the rules
 DECOMPOSITION OF COMPOUND RULES

• A compound rule is a collection of many simple rules

combined together
• Any compound rule structure may be decomposed and
reduced to a number of simple canonical rule forms
• The rules are generally based on natural language
representations
• The rules are:
• I. Multiple conjunctive antecedents:
• IF x is A1, A2, …….An THEN y is Bm
• Let us take Am= 𝐴1 ∩ 𝐴2 ∩. . . . 𝐴𝑛 then
 DECOMPOSITION OF COMPOUND RULES

• Then as we know 𝜇𝐴𝑚 (𝑥) = min{𝜇𝐴1 (𝑥), 𝜇𝐴2 (𝑥), . . . . 𝜇𝐴𝑛 (𝑥)
• The compound rule may be rewritten as
• IF Am THEN Bm
• Multiple disjunctive antecedents
• IF x is A1 OR x is A2 OR……OR x is An THEN y is Bm
• Taking 𝐴𝑚 = 𝐴1 ∪ 𝐴2 ∪. . . .∪ 𝐴𝑛 we have
• IF x is Am THEN y is Bm
• Then we know that
• 𝜇𝐴𝑚 (𝑥) = max{𝜇𝐴1 (𝑥), 𝜇𝐴2 (𝑥), . . . . 𝜇𝐴𝑛 (𝑥)
 DECOMPOSITION OF COMPOUND RULES

• Conditional statements with ELSE and UNLESS

• Example 1: IF A1 THEN(B1 ELSE B2)
• It is equivalent to the decomposition
• IF A1 THEN B1
• OR
• IF NOT A1 THEN B2
• Example 2: IF A1 (THEN B1) (UNLESS A2)
• It is equivalent to the decomposition
• IF A1 THEN B1
• OR
• IF A2 THEN NOT B1
 DECOMPOSITION OF COMPOUND RULES

• Example 3: IF A1 THEN B1 ELSE IF A2 THEN B2

• It can be decomposed into
• IF A1 THEN B1
• OR
• IF NOT A1 AND IF A2 THEN B2
• Example 4: Nested IF- THEN rules
• The rule “ IF A1 THEN (IF A2 THEN B1)”
• Can be interpreted as
• IF A1 and A2 THEN B1
• Based on all the above mentioned methods compound rules
can be decomposed into a series of canonical simple rules
 AGGREGATION OF FUZZY RULES

• Aggregation of rules is the process of obtaining the overall

consequents from the individual consequents provided by
each rule
• Rule 1:
• Conjunctive system of rules
• For a system of rules to be jointly satisfied, the rules are
connected by “and” connectives
• Here, the aggregated output, y, is determined by the fuzzy
intersection of all individual rule consequents, yi, i = 1,2, …n
• y = y1 and y2 and…… and yn
 AGGREGATION OF FUZZY RULES

• Symbolically,
𝑦 = 𝑦1 ∩ 𝑦2 ∩. . . .∩ 𝑦𝑛
The aggregated output can be defined by the membership
function
𝜇𝑦 (𝑥) = min{𝜇𝑦1 (𝑥), 𝜇𝑦2 (𝑥), . . . . 𝜇𝑦𝑛 (𝑥) , ∀𝑥
Disjunctive system of rules:
• In this case the satisfaction of at least one rule is enough. The
rules are connected by “or” connectives.
• Here, the fuzzy union of all individual rule contributions
determine the aggregated output as
 AGGREGATION OF FUZZY RULES

• y = y1 or y2 or…..or yn
• Symbolically,
• 𝑦 = 𝑦1 ∪ 𝑦2 ∪. . . . . .∪ 𝑦𝑛
• Again it can be defined by the membership function
• 𝜇𝑦 (𝑥) = max{𝜇𝑦1 (𝑥), 𝜇𝑦2 (𝑥), . . . . 𝜇𝑦𝑛 (𝑥)
 SOME SUGGESTIONS FOR REVIEW 3

1. Mention whether your project is an innovative one or just a

simulation of existing algorithms.
2. If innovative one then mention the literature review and say
what were the gaps in it.
3. Your approach to fill the gaps.
4. Are you proposing a new algorithm?
5. State the algorithm.
6. Don’t put too much of codes either in the slides or in the
report.
7. Mention the dataset on which it is applied.
8. Mention the measure(s) which are used to explain the
supremacy of your algorithm over the existing ones
 SOME SUGGESTIONS FOR REVIEW 3

• Show the results

• If only simulation of existing systems then show the
characteristics of the system
• Mention the input and output of the system
• Run through an example to illustrate
• NOTE: In both cases show the referred papers
• You can show the web links also
• Can your work be converted into a paper?
• If so, how and why?
• Give me a contact number to interact with me