You are on page 1of 2

Maad M.

Mijwel May 2015

Heuristic Algorithms
Maad M. Mijwel
Computer science, college of science,
University of Baghdad
Baghdad, Iraq
maadalnaimiy@yahoo.com
May 2015

__________________________________________________*****_________________________________________________

design problems, routing problems, scheduling problems,


I. INTRODUCTION social and economic planning problems they are used in many
Operations research is the whole set of methods involving different areas.
finding the most appropriate solution for a given problem.
Mathematical modeling of the real complex problems of the 2 - Swarm Intelligence
world is a scientific field that uses scientific methods such as The swarm algorithms are search algorithms that are inspired
algorithms and statistics. It optimizes the optimal solution for by the movements of the swarms in nature. A lot of
the query. Optimization Problems are the most desirable individuals interact with each other to solve a certain
solutions. In general, NP-Complete is used to solve problems problem. There are a lot of algorithms. Famous among them
such as Decision Problems. are Particle Swarm Optimization, Acritical Bee Colony, Ant
Colony, Firefly Algorithm,
Optimization problems are making your fitness function
minimum or maximum. In other words, increasing 3 - Tabu Search
productivity in a company is maximizing optimization while Solving is prohibited again in the next steps to prevent
reducing the duration of a business in the company is a repetitive movement during the steps leading up. Thus,
minimum optimization. In fact, we face many times in our regional research is conducted to investigate solutions to
daily life with optimism. For example, when I write this achieve the best solution.
article, I try to set up shorter sentences that will give the same
clarification, while optimizing to the minimum, optimizing to 4 - Simulated Annealing
maximize when searching for more and more efficient how to The simulated annealing algorithm can give a reasonable
work today. Optimization problems are classified as discrete approximation for a function with a large search space. The
or continuous optimization according to the structure of place decides between probabilistic survival or reversion to
decision variables in the mathematical model. In some places another situation. This will direct the system to low
you can see intermittent optimization as a combinatorial. energy. Electronic circuit design is widely used in the solution
Interrupted optimization variables include a restricted set. For of problems such as image processing, road finding problems,
example, the famous Traveling Salesman Problem is solved travel problems.
with intermittent optimization. Continuous optimization
problems are problems that do not limit the variables. 5 - Artificial Neural Networks [ANNs]
Artificial neural networks are very functional models for
Intuitive Optimization Algorithms find a place to find pattern recognition and machine learning, which categorize
solutions to their Optimization Problems in operational new patterns from acquired training data. It was inspired by
researches. A real life question that needs to be solved must the neuron function in the animals' brains. Many areas such as
first be cast into a mathematical model. Thus, using this speech analysis, image processing, etc. are used.
mathematical formula as a criterion aims to find the best
solution in a short time. Heuristic Optimization Algorithms do 6 - Support Vector Machines
not guarantee the best solution. The algorithm is considered to Supporter Vector Machine is a model that grasps patterns and
be so effective as to how quickly it reaches a good solution. analyzes data by using training data with artificial
intelligence. These algorithms are used for regression analysis
and classification purposes. Using the sample data, the
II. EXAMPLE ALGORITHMS algorithm will sort the new samples into groups. These SVMs
are a subset of the artificial intelligence that systems learn
from knowledge in machine learning, and require training data
1 - Genetic Algorithms
before analyzing new examples.
Genetic algorithms are subcategories of evolutionary
algorithms. It is inspired by the theory of evolution, with
features such as mutation, crossover, and selection. Parameter III. SAMPLE PROBLEMS
and system diagnosis, control systems, robot applications,
image and voice recognition, engineering designs, planning, 1 - Traveling Salesmen Problem
artificial intelligence applications, expert systems, function Among the cities with distances from each other, the traveler
and combinatorial optimization problems such as network is the problem of solving the minimum distance provided that
all the cities have to pass through once.
1
maadalnaimiy@yahoo.com _____________________________________
Maad M. Mijwel May 2015

2 - N-Queen Problem 500 degrees by subtracting 500 degrees from our problem
A chess board is the problem of being stowed on the board in equation. We have to use absolute value because we need to
such a way that there are no points of attack on a veterinarian get clear distance of 500 den large or small. Here you will
until N. benefit from formulating between 0 and 1 to make the
eligibility value more reducible. I formulated it as 1 / (1 +
IV. STUDY LOGIC Conclusion).
Let's talk a little bit about the rationale. As you know from the 1-) 10 2 +15 2 + -6 3 = 500 | 500-109 | = 391 = 1 / (1 +
Heuristic Optimization Algorithms, we aim to find the best 391) = 0.0025
solution in the shortest possible time from the probing search 2-) -25 2 +29 2 -10 3 = 500 | 500-466 | = 34 = 1 / (1 + 34)
space available. Now let's come to probing. Let's say we are = 0.028
going to place 3 different volume shapes in a box with a 3-) 10 2 +10 2 +4 3 = 500 | 500-264 | = 236 = 1 / (1 +
volume of 500.Let's write some formulas. The first object 236) = 0.0042
occupies the space of the last object cube in the same way as After finding these values, search algorithms try to reach
the square of the lowest object, in the second object. We can better solution clusters by applying special processes to
take a mathematical model such as x 1 2 + x2 2 + x 3 2 = them. Because of the form we create here, it returns 1 in the
500 and try to find the values of x 1 , x 2 and x 3 in best solution set.
the problem space in the search space. Search space x 1 , x 2, I gave the example above to understand the working logic of
x 3 represents the maximum and minimum value range of the most of the heuristic optimization algorithms. We randomly
parameter values. Setting a limit on the parameters we will use set the solution clusters in the search space at startup. The
in the equation of the problem at hand will speed up the math function I was building was purely exemplary. There are
solution reach. If the limit is not set, an infinite number of also mathematical functions to test the performance of
clusters will appear for all parameters. The more the search heuristic optimization algorithms.
space is limited, the faster we will go. To restrict search space, In the next installment, you can reinforce what is described by
we must determine the minimum and maximum values that looking at examples of search algorithms.
each parameter can take. So we can approach solving As an example of heuristic optimization algorithms;
faster. Of course, search space should be carefully determined Genetic Algorithm (GA)
by trying not to exclude values that might be a better solution Ant Colony Optimization (ACO)
from the search space. You must determine the parameter Particle Swarm Optimization (PSO)
value ranges according to your probing. I have problem x 1, Artificial Bee Colony (ABC)
x 2 , xI set this limit by thinking that the values from -30 to 30 Differential Evolution Algorithm (DEA)
could produce a solution. Also, when solving the problem, let's Simulated Annealing (SA)
assume that we have formulated our fitness function so that Gravity Search Algorithm (GSA)
there is no space in the box. In this case, the rest of the box is Gas Brownian Motion Optimization (GBMO)
meant to be minimized. Initially, random solution sets are Heat transfer search (HTS)
created and the fitness values of the resulting solution sets are Electromagnetic Field Optimization (EFO)
found. Let's create a random set of 3 solutions: Optic Inspired Optimization (OIO)
1-) x 1 = 10, x 2 = 15, x 3 = -6 Weighted Superposition Attraction (WSA)
2-) x 1 = -25, x 2 = 29, x 3 = 10 Forest Optimization Algorithm (FOA)
3-) x 1 = 10, x 2 = 10, x 3 = 4. Hurricane Based Optimization Algorithm (Hurricane Based
Now we must calculate the fitness values of these solution Optimization Algorithm)
clusters. The suitability value is determined by intention Firefly Algorithm (Firefly algorithm) (FA)
parameters and their suitability for solution of the problem is Tree-Seed Algorithm (TSA)
examined. By replacing the values, I will check the distance of

2
maadalnaimiy@yahoo.com _____________________________________

You might also like