Information Technology Reference
In-Depth Information
Artificial Bee Colony (ABC) Algorithm is a novel swarm intelligence based
algorithm introduced by Karaboga and Basturk to solve the optimization problem of
multivariable functions [4]. Compared with Genetic Algorithm (GA) [5], Particle
Swarm Optimization (PSO) [6] or some other population based algorithms, ABC
algorithm has the faster convergence speed and the better ability to get out of a local
optimal solution [4]. As a consequence, ABC algorithm attracts extensive attention
and obtains rapid development in various fields.
Although ABC algorithm has good performances on optimization, there are still
some insufficiencies in the selection strategy. In the onlooker bee stage of traditional
ABC algorithm, an onlooker randomly chooses a food source with the probability
value calculated from roulette wheel selection (RWS) strategy. However, compared
with other selection strategies, RWS has two key disadvantages. Firstly, although
RWS is a well-known random selection strategy according to the proportion of
different components of a system, it is inconvenient when solving minimization
problems because the fitness values have to be converted to make the smaller value
has more probability to be chosen. Secondly, RWS has no adjustable parameters to
adapt its selection pressure to different optimization problems. In order to improve the
optimization ability of ABC algorithm, these two disadvantages should be avoided. In
this paper, we propose to adopt tournament selection (TS) strategy in ABC to avoid
the above two disadvantages. In the literatures, TS has been used in some
optimization algorithms, such as Genetic Algorithm (GA), that shows good
performance in population selection [7]. TS strategy can be applied easily in both
minimization and maximization problems because it compares the fitness values to
select better solutions. Moreover, TS strategy has a parameter
(the proportion of
solutions to be chosen) that can adjust selection pressure for various optimization
problems. Therefore, the proposed TS based ABC (TSABC) algorithm can extend the
generality of ABC algorithm. Another contribution of this paper is to design the elitist
strategy to the standard ABC framework. Although ABC has been widely studied, it
is strange that no literatures specifically claim that the ABC algorithm should not
abandon the historically best solution that has found so far. On scout stage of ABC,
the food source (solution) that cannot be improved for a certain generations will be
abandoned. However, if the solution is the globally best one of the population, such
abandon can make the ABC algorithm deteriorated. In order to avoid such bad
influence on optimization, we propose the elitist strategy to ABC. This strategy uses
an extensive archive to keep the historically best solution during the running. This
solution is not in the ABC population, but is updated in every generation if the
globally best solution of the population is better than this solution. Therefore, even
though the scout stage may abandon the globally best solution of the population (e.g.,
it has not improved for a long time), the historically best solution won't be
abandoned. This is only a slight change to the ABC framework. It not only has no
negative influences on the performance, but also can be applied to any ABC variants,
which can be considered as a standard component in the ABC algorithmic framework.
This paper aims to improve ABC algorithm with TS strategy and compare its
performance with that with RWS. Also, the optimization ability of TSABC algorithm
is analyzed under the change of control parameter values. The rest of the paper is
structured as follows. Section 2 introduces ABC algorithm and its current
development. The TSABC algorithm and the experimental results are presented in
Section 3 and Section 4, respectively. Finally, conclusions are given in Section 5.
ʻ
Search WWH ::
Custom Search