0% found this document useful (0 votes)
105 views5 pages

An Overview of Genetic Algorithms A Structural Analysis

The Genetic Algorithm (GA) is a genetic and natural selection-based search-based optimization technique.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
105 views5 pages

An Overview of Genetic Algorithms A Structural Analysis

The Genetic Algorithm (GA) is a genetic and natural selection-based search-based optimization technique.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Volume 6, Issue 5, May – 2021 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

An Overview of Genetic Algorithms:


A Structural Analysis
Satish Kumar Gupta
Assistant Professor, Department of Information Technology,
Gopal Narayan Singh University, Jamuhar Sasaram (Rohtas), Bihar-821305

Abstract:- The Genetic Algorithm (GA) is a genetic and since been successfully evaluated on a variety of optimization
natural selection-based search-based optimization problems.
technique. It is often employed in the search for optimal
or near-optimal solutions to complex problems that could In GAs, we have a population or reservoir of potential
take a lifetime to solve. It's widely used in science and solutions to a crisis. Recombination and mutation are then
machine learning to solve optimization problems. In applied to these solutions (as in natural genetics), resulting in
computer science and operations research, a genetic the birth of For decades, the technique has been used. A
algorithm (GA) is a metaheuristic influenced by natural fitness value is given to each individual (or candidate
selection that belongs to the broader class of evolutionary solution), (based on the value of its objective function), and
algorithms. The aim of this article is to look about the the fitter it is, the better ones have a better chance of mating
Genetic Algorithm, Basic Terminologies and the it’s and producing more “fitter” individuals. This is consistent
various applications in different fields. This paper also with Darwin's "Survival of the Fittest" theory.
focuses on explaining the various types of representation In this manner, we continue to "evolve" Before we
of Genetic Algorithm and various Merits and Demerits of reach a point where we must quit, we must develop great
Genetic Algorithm (GA). people or alternatives over years.

Keywords:- Algorithm, Operator, Terminology, Presentation. Despite being sufficiently random, genetic algorithms
outperform random local quest (in which we simply pursue
I. HISTORY various random solutions) and keep track of the best so far)
since they also use historical data.
A. Rechenberg introduced evolutionary computing in his
work "Evolution strategies" in the 1960s (Evolutions Genetic Algorithm:
strategies in original). Other researchers expanded on his “Genetic Algorithms are good at exploring big,
definition. John Holland invented Genetic Algorithms potentially enormous search spaces, searching for ideal
(GAs), which he and his students and colleagues created. combinations of items that you may not find in a lifetime.”
B. John Koza used a genetic algorithm to evolve algorithms
to perform specific tasks in 1992. His system was dubbed Basic Terminology:
"genetic programming" by him (GP). Since LISP Before diving into the topic of Genetic Algorithms, it's
programs can be represented in a "parse tree," which is the important to understand some of the vocabulary that will be
object the GA deals with, they were used. used in this tutorial.

II. INTRODUCTION  Population: The algorithm begins with a population of


solutions (represented by chromosomes). A population's
The Genetic Algorithm (GA) is genetics and natural Solutions are sought and implemented in order to establish
selection-based search-based optimization technique. It has a new population. This is driven by a desire for the new
often been employed to quest for ideal or near-optimal population to be superior to the existing. Solutions chosen
solutions to complex problems that will take an eternity to to form new solutions (offspring) are chosen based on
solve otherwise It's a popular tool for optimization, analysis, their fitness; the better they are, the more likely they are to
and machine learning. replicate.
 Chromosomes: Cells are the building blocks of all living
Nature has always been a source of great inspiration for organisms. The chromosomes in each cell are identical.
all of humanity. GAs is search-oriented algorithms that are Chromosomes are DNA strings that function as a
based on natural selection and genetics concepts. GAs is a blueprint for the entire organism. A chromosome is made
subset of Evolutionary Computation, which is a much larger up of chromosomes, which are DNA blocks. Each gene
computing branch. codes for a specific protein. Essentially, each gene codes
GAs was invented at the University of Michigan. The for a characteristic, such as the color of one's eyes. Alleles
algorithms were developed by John Holland and his students are the variouscharacter's settings (for example, blue and
and colleagues, most notably David E. Goldberg, and have brown). Each gene is found on a different chromosome.

IJISRT21MAY1057 www.ijisrt.com 1305


Volume 6, Issue 5, May – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
This is a challenging job opportunity is referred to as
locus.
 Gene: Potential theories in a GA are made up of
chromosomes made up of genes in turn. In a GA,
chromosomes are usually represented as binary strings, a
series of 1s and 0s that denote the inclusion or exclusion
of specific items based on their position in the string.
Within such a chromosome, a gene is a single piece.
 Allele: It is the weight a gene has on a specific
chromosome.  Fitness Function: A fitness function takes a solution as
an input and outputs its suitability. The health and
objective functions can overlap in some cases are the
same; however, depending on the problem, they may be
different.
 Genetic Operators: Genetic Operators affect the
offspring's genetic structure. This includes things like
crossbreeding, mutation, and selection.

III. BASIC STRUCTURE

A GA's basic structure is as follows:


 Genotype: In the computation space, the population is We begin by selecting parents for mating from an initial
known as genotype. The solutions are So that a computing population (which may be produced randomly or seeded by
machine can efficiently interpret and control them, they other heuristics). To create new offspring, on the kin, use
are represented in the computation space. crossover and mutation operators. Finally, these offspring
 Phenotype: The community in the actual real-world replace the new individuals in the population, and the cycle
solution space, where solutions are described exactly as begins again. Genetic algorithms aim to approximate human
they are in the model the actual world conditions, is evolution in this manner to some extent.
referred to as phenotype.
 Parent: We will have a population of candidate solutions
at any point in the algorithm. Some candidate solutions
will be chosen in each iteration, or generation, of the
algorithm to survive into the next generation and produce
some children.
 Children: Selected parents from the previous generation
have created new candidate solutions.
 Decoding and Encoding: Basic problems have the same
phenotype and genotype spaces. The phenotype and
genotype spaces, on the other hand, are rarely the same.
Encoding is transforming a solution from genotype to
phenotype space, while decoding is transforming a
solution from genotype to phenotype space, the
transformation from the space of phenotype to the space
of genotype since encoding occurs several times in a GA
during the computation of fitness values, it should be fast.
How they Work:
Consider the 0/1 Knapsack Problem, as an illustration They're primarily interested in the natural selection
The Phenotype space is made up of only the component process, which means they extend natural selection's basic
numbers of the objects to be chosen. properties to whatever issue we're attempting to solve.

It can be represented as an n-byte binary string in A genetic algorithm basic procedure is as follows:
genotype space, however (where n is the number of items). 1. Initialization: Make a starting population. This
The next object is selected when a 0 appears at position x, population is typically produced at random and can range
while a 1 indicates the opposite. In this case, the genotype in size from a few individuals to thousands.
and phenotype spaces are not the same. 2. Evaluation: After that, we test. Each person in the
population is given a "fitness" ranking. The importance of
exercise is measured by how well it does our ideal criteria.
These specifications may be straightforward, such as
"faster algorithms are better," or more complicated, such

IJISRT21MAY1057 www.ijisrt.com 1306


Volume 6, Issue 5, May – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
as "Sturdier materials are preferable, but they should not
be too so."
3. Selection: We want to continue to improve our
population's physical health. We can achieve this with the
help of filtering, which removes the poor designs from the
population and leaves only the best individuals. There are
many selection approaches, but the core idea remains the
same: make it more likely that fitter individuals will be
Genotype Representation:
selected for our next generation.
The representation we will use to represent our
4. Crossover: We merge components of our selected
solutions when applying a genetic algorithm is one of the
individuals to create new individuals during the crossover.
most crucial choices to make. It has been discovered that poor
This is comparable to how sex behaves in nature. The
representation can result in poor GA results.
hope is that by combining these traits from two or more
people, we will be able to create a more 'fit' offspring who
As a result, picking the suitable representation and
will inherit the best traits from both parents.
defining the mappings between the phenotype and genotype
5. Mutation: We ought to inject some randomness into the
spaces is critical to a GA's success.
genetics of our populations, or else any possible answer
will end up in our initial population. Mutation usually
Binary Representation:
works by making minor changes to a person's genome at
In GAs, this is one of the most basic and commonly
random.
used representations. The genotype in this form of
6. And repeat: Now that we have our next generation, we
representation is made up of bit strings.
can go back to phase two and repeat the process before
reaching a state of termination.
The binary representation is normal for complex
7. Termination:
problems where the solution space consists of Boolean
You may want to halt the genetic algorithms search for
decision variables – yes or no. Think about the 0/1 Knapsack
a solution for a variety of reasons. The possible explanation is
Problem. If there are n products, a solution can be expressed
that algorithm has discovered a satisfactory solution and
as a binary string of n elements, with the xth element
satisfies a set of minimum requirements. Constraints such as
indicating whether item x is selected (1) or not (0).
time or money may be offered as grounds for terminating.

Operators of Genetic Algorithms:


Following the development of the initial generation, the
algorithm evolves the generation using the following
The binary representation of numbers can be used to
operators –
describe them in other problems, especially those involving
1) Selection Operator: The goal is to give people who have
numbers. Since various bits have different definitions in this
poor health scores a leg up to inspire them to keep going pass
form of encoding, mutation and crossover operators can be
on their genes to future generations.
ineffective unexpected results. Gray coding can help with this
2) Crossover Operator: This describes person mating. The
to some degree, as a change of one bit has a minor effect on
crossover sites are chosen at random, and two individuals are
the solution.
chosen using the selection operator. The genes at these
crossover sites are then swapped, resulting in creating an
Real-Valued Representation:
entirely new body (offspring).
The real-valued representation is the most natural when
Rather than discrete variables, we'd like to describe genes
For instance-
using continuous variables. These genuine or floating-point
numbers, however, are limited by the computer's precision.

Integer Representation:
We can't limit the solution space for discrete-valued
genes to binary "yes" or "no" forever. E.g., the four distances
can be encoded as 0,1,2,3 if we want to encode North, South,
East, and West. In such instances, it is preferable to use
3) Mutation Operator: The central idea is to insert random
integer representation.
genes into embryos to maintain population variability and
avoid premature fusion. For instance –

IJISRT21MAY1057 www.ijisrt.com 1307


Volume 6, Issue 5, May – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Permutation Representation: Fluctuating Population:
The solution to an order of elements is used to express In a generational model, we produce "n" heirs, where n
several problems. Permutation representation is the best is the population size, and after each cycle, the whole
choice in these contexts. population is replaced.

The travelling salesman problem is a well-known Application of the Genetic Algorithm:


example of this kind of representation (TSP). Before Genetic algorithms are often used in some applications,
returning to the starting city, the seller must tour all cities, but they are most widely used in optimization problems of
visiting each one exactly once.The tour's overall distance various kinds.
must be kept to a minimum. Since the TSP's solution is a
standard ordering or permutation of all the towns, it makes  Genetic Algorithms for Optimization: When solving
sense to use a permutation representation for this issue. optimization problems, we must optimise or minimise the
value of an objective function, genetic algorithms are most
commonly used while adhering to a set of constraints.
Throughout the tutorial, the approach to solving
Optimization problems has been illustrated.
IV. POPULATION  In Economics: Models like the cobweb model, game
theory equilibrium resolution, and asset pricing are all
The population is now a subset of solutions in the described using Gas, and so on.
modern era. It is also known as a chromosome set. There are  Recurrent Neural Networks: GAs is also used to train
a couple of things to keep in mind when dealing with GA recurrent neural networks.
populations:  Image Processing: GAs is also used for dense pixel
• It is critical to preserve demographic diversity, as premature matching in digital image processing (DIP).
convergence will result.  Vehicle Routing Problems − with a diverse fleet,
• A large population can cause a GA to slow down, and a numerous soft time windows, and multiple depots
small population cannot have a sufficient mating pool. As a  GAs is also used to solve various Scheduling Problems,
result, trial and error must be used to decide the best especially the timetabling problem, in scheduling
population scale. A population is typically described using a applications.
two-dimensional array of population size, size x, and
 Machine Learning GBML (Genetics-Based Machine
chromosome size.
Learning) is a niche field of machine learning, as
previously stated.
Population Initialization:
 Robot Trajectory Generation: GAs is used to plan a
In a GA, there are two critical methods for initializing a
robot arm's path from one point to the next.
population.
• Random Initialization: To populate the initial population,  Parametric Design of Aircraft: By varying criteria and
findings, GAs has been used to manufacture aircraft.
completely random solutions are used.
• Heuristic initialization: Using a known heuristic for the  Travelling Salesman Problem and its Solutions: Using
dilemma, populate the initial population. novel crossover and packaging techniques, the TSP, a
well-known combinatorial puzzle, has been solved using
It has been discovered that using a heuristic to initialize GAs.
the entire population results in a population with identical
solutions and no diversity. Random solutions have been Advantages of Genetic Algorithm:
discovered to be the ones that move the population to GAs has several advantages that have helped them
optimality. As a consequence, rather than flooding the become highly successful.
population with heuristic-based solutions, we seed it with a a. It is impractical to use derivative knowledge (which might
diverse set of solutions and fill in the gaps with random not be usable for certain real-world problems).
solutions using heuristic initialization. b. As opposed to conventional methods, it is quicker and
more effective.
It was also discovered that heuristic initialization only c. Optimizes both continuous and discrete functions, as well
influences the population's initial fitness in certain situations; as multi-objective problems, thanks to its solid, parallel
Optimality is aided by the variety of options. capabilities.
d. Rather than providing a single "good" alternative, it
Population Models: provides a list of "good" choices.
The steady-state and fluctuating demographic models e. Gets a solution to the problem often and it gets better over
are the two most popular population models. time.
f. Useful when there are several parameters to consider and
Steady-State: the search space is enormous.
Each version of steady-state GA produces one or two
descendants, each replacing one or two individuals in the
population. Stable condition GA is also known as incremental
GA.

IJISRT21MAY1057 www.ijisrt.com 1308


Volume 6, Issue 5, May – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
V. LIMITATIONS OF GENETIC ALGORITHM REFERENCES

While genetic algorithms have proven to be a quick and [1]. Azadivar, F. and Tompkins, G., Simulation
efficient problem-solving method, they do have some optimization with qualitative variables and structural
disadvantages. Any of these flaws will be discussed later. model changes: a genetic algorithm approach. Euro. J.
1. Determining a representation for the problem is the first Opt. Res., 1999, 113(1), 169–183.
and most critical factor in developing a genetic algorithm. [2]. Chan, K.C., and Tansri, H., A study of genetic
The terminology used to specify candidate solutions must crossover operations on the facilities layout
be robust because it must be able to withstand random problem. Computers & Ind. Eng., 1994, 26(3), 537–550.
changes without causing fatal errors. [3]. Chen, C.L., Vempati, V.S. and Aljaber, N., An
2. Coding the fitness (evaluation) function to achieve higher application of genetic algorithms for flow shop
fitness and provide better solutions to the problem at hand problems. Euro. J. Opt. Res., 1995, 80(2), 389–397.
is one of the most challenging genetic algorithms. A poor [4]. Meyer, L., Feng, X.: A fuzzy stop criterion for genetic
fitness function choice can result in serious problems, algorithms using performance estimation. In:
such as being unable to solve a problem or, even worse, Proceedings of the Third IEEE Conference on Fuzzy
returning the incorrect solution to a problem. Systems. (1994) 1990–1995
3. In addition to choosing a good fitness function, a Genetic [5]. Howell, M., Gordon, T., Brandao, F.: Genetic learning
Algorithm's other parameters, such as population size, automata for function optimization. IEEE Transactions
mutation rate, and crossover rate, must be carefully on Systems, Man, and Cybernetics-Part B: Cybernetics
chosen. The Genetic Algorithm would not be able to 32 (2002) 804–815
achieve accurate results with a small population size [6]. Poli, R.: Exact schema theorem and effective fitness for
because there would not be enough solution space. A high GP with one-point crossover. In Whitley, L.D.,
rate of genetic change, or a faulty selection system, will Goldberg, D.E., Cant´u-Paz, E., Spector, L., Parmee,
destroy the beneficial schema, and the population will hit I.C., Beyer, H.G., eds.: Proceedings of the Genetic and
error catastrophe, evolving too rapidly for selection ever Evolutionary Computation Conference, Morgan
to achieve convergence. Kaufmann (2000) 469–476
4. Computational problems cannot be solved with genetic [7]. De Jong, K.A.: Genetic algorithms are NOT function
algorithms. Although genetic algorithms can find accurate optimizers. In Whitley, L.D., ed.: Proceedings of the
solutions to these problems, traditional analytic methods Second Workshop on Foundations of Genetic
can faster and with fewer computational steps. Algorithms, Morgan Kaufmann (1993) 5–17
5. When developing Generic algorithm solutions, GA [8]. B¨ack, T., Hammel, U., Schwefel, H.P.: Evolutionary
scientists must also consider premature convergence. An computation: Comments on the history and current
individual could be physically fitter than any of its state. IEEE Transactions on Evolutionary Computation
competitors. As a result, rather than searching the fitness 1 (1997) 3–17
landscape extensively enough to find the global optimum, [9]. Burkowski FJ (1999) Shuffle crossover and mutual
this individual may replicate many more new individuals, information. Proceedings of the 1999 Congress on
reducing the diversity of the new population and causing Evolutionary Computation-CEC99 (Cat. No.
the algorithm to converge on the local optimum that 99TH8406), Washington, DC, USA, 1999, pp. 1574–
represents that specific individual. This type of 1580
inefficiency is most common in small problems, where [10]. Deb K, Agrawal RB (1995) Simulated binary crossover
even minor differences in reproduction rate can lead to for continuous search space. Complex Systems 9:115–
one genotype dominating the others. 148
Websites:
VI. CONCLUSIONS [1]. https://www.tutorialspoint.com/genetic_algorithms/inde
x.htm
The richness of genetic algorithms and the various [2]. Darrell Whitley, Computer Science Department
parallel models is one of the most striking features of this Colorado State Universit, Fort Collins CO
kind of computing. Minor changes to the algorithm will often [3]. https://www.cs.jhu.edu/~ayuille/courses/Stat202C-
result in unpredictable emergent behaviour. Our Spring10/ga_tutorial.pdf
understanding of genetic algorithms has also improved due to [4]. https://www.researchgate.net/post/How_can_I_decide_t
recent scientific advances, allowing us to use more he_stopping_criteria_in_Genetic_Algorithm
sophisticated analytical methods. [5]. https://www.cs.rochester.edu/u/nelson/courses/csc_173/
genetic-algs/algorithm.html
After study of this article, readers are well known about [6]. https://www.sciencedirect.com/topics/engineering/genet
the how many Terminologies are used for GA, How the GA ic-algorithm
Model working and what are the Advantages and [7]. https://www.geeksforgeeks.org/genetic-algorithms/
Disadvantages of GA. This Article also addresses a wide [8]. https://ocw.mit.edu/courses/institute-for-data-systems-
range of current issues. and-society/ids-338j-multidisciplinary-system-design-
optimization-spring-2010/lecture-
notes/MITESD_77S10_lec11.pdf
[9]. http://pages.cs.wisc.edu/~dyer/cs540/notes/ga.html

IJISRT21MAY1057 www.ijisrt.com 1309

You might also like