Scribd
Upload
115 views2 pages

Comparison Between Genetic Algorithms and Particle Swarm Optimization

Uploaded by

GNANAPRAKASH
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
115 views2 pages

Comparison Between Genetic Algorithms and Particle Swarm Optimization

Uploaded by

GNANAPRAKASH
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/ 2

Comparison between Genetic Algorithms and Particle

Swarm Optimization
R u s s e l l C. E b e r h a r t and Y u h u i Shi
Department of Electrical Engineering
Indiana University Purdue University Indianapolis
723 W. Michigan St., SL160
Indianapolis, IN 46202
eberhart,[email protected]

Abstract
This paper compares two evolutionary computation paradigms: genetic algorithms and particle
swarm optimization. The operators of each paradigm are reviewed, focusing on how each affects
search behavior in the problem space. The goals of the paper are to provide additional insights
into how each paradigm works, and to suggest ways in which performance might be improved
by incorporating features from one paradigm into the other.

Introduction
Four well-known paradigms currently exist in evolutionary computation: genetic
algorithms [5], evolutionary programming [4], evolution strategies [9], and genetic
programming [8]. A new evolutionary computation technique, called particle swarm
optimization (PSO), inspired by social behavior simulation, was originally designed
and developed by Eberhart and Kennedy [2,3,6,7]. In PSO, instead of using more
traditional genetic operators, each particle (individual) adjusts its "flying" according to
its own flying experience and its companions' flying experience.

This paper compares genetic algorithms and particle swarm optimization. Operators
that are used by each paradigm are reviewed. The focus is on how each operator affects
the paradigm's behavior in the problem space.

There are, of course, many ways to implement a genetic algorithm (GA). Regardless of
the specific implementation, it is generally agreed that GAs utilize one form or another
of three operators: selection, crossover, and mutation. We will examine
implementations of these operators, and compare them with PSO operators. In this
paper, it can be assumed that, in most cases, a basic, binary version of a GA is being
referred to, such as the "plain vanilla" GA in [2] or the elementary GA 1-1 in [1].

The authors readily concede that few applications use these basic GA configurations,
and a number of modifications to these configurations are also examined. It should be
noted that it is not the goal of this paper to compare GAs and PSO in order to declare
one or the other as somehow better. Rather, the goals are to provide insights into how
GAs and PSO work, and to suggest ways in which performance might be enhanced by
incorporating features from one paradigm into the other.
612

Analysis
In PSO, a particle is analogous to a population member (chromosome) in a GA. Like a
GA chromosome, a particle represents a candidate solution to the problem being
addressed. Each particle is treated as a point in the D-dimensional problem space. The
ith particle is represented as X~ = (Xil,Xi2. . . . . Xio). The best previous position (the
position giving the best fitness value) of the ith particle is recorded and represented as
P1 = (Pil,Pi2, . . . . PiD). The index of the best particle among all the particles in the
population is represented by the symbol g. The rate of the position change (velocity)
for particle i is represented as VI = (vil, vi2. . . . . ViD). The particles are manipulated
according to the following equations:

via=w*via+cl *randO*(pid-xia)+c2 *RandO*(pga-Xia) (I)


Xid = Xid + Vid (2)

By adding the inertia weight w into PSO, a new version of PSO is introduced in [10].
The inertia weight w is employed to control the impact of the previous history of
velocities on the current velocity, thereby influencing the trade-off between global
(wide-ranging) and local (fine-grained) exploration abilities of the "flying points." A
larger inertia weight facilitates global exploration (searching new areas) while a
smaller inertia weight tends to facilitate local exploration to free-tune the current
search area. Suitable selection of the inertia weight provides a balance between global
and local exploration abilities and thus requires fewer iterations on average to find the
optimum [10]. (In some ways, the name damping weight might be more descriptive,
but the term inertia weight is used in this paper.)

In a GA, each of the three main classes of operations (selection, crossover, and
mutation) can be implemented in a number of ways. PSO does not label its operations
in the same way as GAs, but analogies exist. These analogies depend, of course, on the
implementation of the GA operation. Complicating any comparisons is the fact that the
effects of the various operations often vary over the course of a run. (A run is defmed
as the total number of generations of the GA prior to termination. Termination usually
occurs either because a prescribed fitness level has been achieved by at least one
member of the population, or because the maximum number of generations allowed
has been reached.)

Effects of GA crossover, for example, usually vary significantly during a run. At the
beginning, the population members are usually randomized, so that crossover can have
significant effects, moving a chromosome a relatively large distance in problem space.
Toward the end of a run, a population has often converged, meaning that many, if not
most, chromosomes have similar structures. Crossover then usually has less effect, and
the resulting movements are relatively smaller. Adding another layer of complexity is
the fact that the probability of crossover is sometimes varied during a run, often
starting out at with a relatively large probability, and ending with a smaller one.

PSO does not have a crossover operation. However, the concept of crossover is
represented in PSO because each particle is stochastically accelerated toward its own
previous best position, as well as toward the global best position (of the entire

You might also like