Page 31 - Fister jr., Iztok, and Andrej Brodnik (eds.). StuCoSReC. Proceedings of the 2015 2nd Student Computer Science Research Conference. Koper: University of Primorska Press, 2015
P. 31
ards the Development of a Parameter-free Bat
Algorithm
Iztok Fister Jr., Iztok Fister Xin-She Yang
University of Maribor Middlesex University
Faculty of Electrical Engineering and Computer School of Science and Technology
Science The Burroughs
Smetanova 17, 2000 Maribor London, United Kingdom
iztok.fister1@um.si stochastic nature-inspired methods are among the promis-
ing kind of optimization algorithms. This family consists
ABSTRACT of the following algorithms: genetic algorithms [7], genetic
programming [11], evolution strategies [1], evolutionary pro-
The bat algorithm is a very simple and efficient nature- gramming [5], differential evolution [14], particle swarm op-
inspired algorithm belonging to the swarm intelligence fam- timization [9], firefly algorithm [18], bat algorithm [19] and
ily. Recent studies showed its potential in solving the con- dozens of others [4].
tinuous and discrete optimization problems and many re-
searchers have applied bat algorithm to solve real-world prob- The bat algorithm (BA)is one of the latest algorithms in the
lems. However, empirical studies showed that the main is- swarm intelligence (SI) domain. Due to its simplicity, it is
sue of most algorithms is the setting of control parameters. a very efficient as well. The original BA works with five pa-
An algorithm such as the bat algorithm typically has a few rameters that represent a potential problem for users which
parameters, and it is very time-consuming to find its best usually may not know how to specify their values properly.
parameter combination. Here, we propose a new parameter- Therefore, this paper introduces a new parameter-free or pa-
free bat algorithm variant without the need for control pa- rameterless BA (PLBA) that eliminates this drawback and
rameters. Initial experiments on basic benchmark functions then proposes techniques for a rational and automated pa-
showed the potential of this new approach. rameter setting on behalf of the user. This study is based
on the paper of Lobo and Goldberg [12].
Keywords
The structure of the remainder of the paper is as follows.
bat algorithm, control parameters, optimization Section 2 presents a description of the original BA. In Sec-
tion 3, a design of the parameter-free or parameterless BA
1. INTRODUCTION (also PLBA) is presented. Experiments and results are dis-
cussed in Section 4. The paper concludes in Section 5 where
As the world is becoming a global village, advances in tech- the directions for the future work are also outlined.
nologies means that new challenges are emerging. Better
and greener products are needed for companies to have a 2. BAT ALGORITHM
competitive edge. Thus, the optimization of product designs
and manufacturing processes become ever-more important. The origin of the BA development can date back to the
It is expected that artificial intelligence is among the most year of 2010 when Xin-She Yang in [19] proposed the new
promising developments for the next 20 years. Many artifi- nature-inspired algorithm that mimics the phenomenon of
cial intelligence tasks can be achieved by optimization and echolocation by some species of bats for the optimization
machine learning techniques. process. This algorithm can integrate the characteristics of
many algorithms such as particle swarm optimization (PSO)
One of the most important tasks for many productions is algorithm [9] and simulated annealing (SA) [10]. Primarily,
how to improve the production of products and services for BA was developed for continuous optimization, but many re-
the market, which requires the optimal design and the op- cent papers showed the potential of the algorithm in solving
timal use of resources. Until recently, many methods have discrete optimization problems. In a nutshell, this algorithm
been used to help designers and developers to solve such op- is widespread in many areas of optimization and industrial
timization problems. Some methods are pure mathematical, applications [8,13,15–17]. Yang proposed the following rules
while others are combined with computer sciences. In line which mimics the bat behavior:
with this, optimization algorithms play a big role. Recently,
• All bats use echolocation to sense the distance to target
objects.
• Bats fly with the velocity vi at position xi, the fre-
quency Qi ∈ [Qmin, Qmax] (also the wavelength λi),
the rate of pulse emission ri ∈ [0, 1], and the loudness
StuCoSReC Proceedings of the 2015 2nd Student Computer Science Research Conference 31
Ljubljana, Slovenia, 6 October
Algorithm
Iztok Fister Jr., Iztok Fister Xin-She Yang
University of Maribor Middlesex University
Faculty of Electrical Engineering and Computer School of Science and Technology
Science The Burroughs
Smetanova 17, 2000 Maribor London, United Kingdom
iztok.fister1@um.si stochastic nature-inspired methods are among the promis-
ing kind of optimization algorithms. This family consists
ABSTRACT of the following algorithms: genetic algorithms [7], genetic
programming [11], evolution strategies [1], evolutionary pro-
The bat algorithm is a very simple and efficient nature- gramming [5], differential evolution [14], particle swarm op-
inspired algorithm belonging to the swarm intelligence fam- timization [9], firefly algorithm [18], bat algorithm [19] and
ily. Recent studies showed its potential in solving the con- dozens of others [4].
tinuous and discrete optimization problems and many re-
searchers have applied bat algorithm to solve real-world prob- The bat algorithm (BA)is one of the latest algorithms in the
lems. However, empirical studies showed that the main is- swarm intelligence (SI) domain. Due to its simplicity, it is
sue of most algorithms is the setting of control parameters. a very efficient as well. The original BA works with five pa-
An algorithm such as the bat algorithm typically has a few rameters that represent a potential problem for users which
parameters, and it is very time-consuming to find its best usually may not know how to specify their values properly.
parameter combination. Here, we propose a new parameter- Therefore, this paper introduces a new parameter-free or pa-
free bat algorithm variant without the need for control pa- rameterless BA (PLBA) that eliminates this drawback and
rameters. Initial experiments on basic benchmark functions then proposes techniques for a rational and automated pa-
showed the potential of this new approach. rameter setting on behalf of the user. This study is based
on the paper of Lobo and Goldberg [12].
Keywords
The structure of the remainder of the paper is as follows.
bat algorithm, control parameters, optimization Section 2 presents a description of the original BA. In Sec-
tion 3, a design of the parameter-free or parameterless BA
1. INTRODUCTION (also PLBA) is presented. Experiments and results are dis-
cussed in Section 4. The paper concludes in Section 5 where
As the world is becoming a global village, advances in tech- the directions for the future work are also outlined.
nologies means that new challenges are emerging. Better
and greener products are needed for companies to have a 2. BAT ALGORITHM
competitive edge. Thus, the optimization of product designs
and manufacturing processes become ever-more important. The origin of the BA development can date back to the
It is expected that artificial intelligence is among the most year of 2010 when Xin-She Yang in [19] proposed the new
promising developments for the next 20 years. Many artifi- nature-inspired algorithm that mimics the phenomenon of
cial intelligence tasks can be achieved by optimization and echolocation by some species of bats for the optimization
machine learning techniques. process. This algorithm can integrate the characteristics of
many algorithms such as particle swarm optimization (PSO)
One of the most important tasks for many productions is algorithm [9] and simulated annealing (SA) [10]. Primarily,
how to improve the production of products and services for BA was developed for continuous optimization, but many re-
the market, which requires the optimal design and the op- cent papers showed the potential of the algorithm in solving
timal use of resources. Until recently, many methods have discrete optimization problems. In a nutshell, this algorithm
been used to help designers and developers to solve such op- is widespread in many areas of optimization and industrial
timization problems. Some methods are pure mathematical, applications [8,13,15–17]. Yang proposed the following rules
while others are combined with computer sciences. In line which mimics the bat behavior:
with this, optimization algorithms play a big role. Recently,
• All bats use echolocation to sense the distance to target
objects.
• Bats fly with the velocity vi at position xi, the fre-
quency Qi ∈ [Qmin, Qmax] (also the wavelength λi),
the rate of pulse emission ri ∈ [0, 1], and the loudness
StuCoSReC Proceedings of the 2015 2nd Student Computer Science Research Conference 31
Ljubljana, Slovenia, 6 October
