Page 33 - Fister jr., Iztok, and Andrej Brodnik (eds.). StuCoSReC. Proceedings of the 2016 3rd Student Computer Science Research Conference. Koper: University of Primorska Press, 2016
P. 33
ew population-based nature-inspired algorithm every
month: Is the current era coming to the end?
Iztok Fister Jr., Uroš Mlakar, Janez Brest, Iztok Fister
Faculty of Electrical Engineering and Computer Science, University of Maribor
Smetanova 17, 2000 Maribor, Slovenia
iztok.fister1@um.si
ABSTRACT and software. Thus, people were unable to test and de-
velop their methods widely. However, some years after the
Every month at least one new population-based nature-insp- creation of artificial neural networks, another discipline (or
ired algorithm has been released in literature. Until recently, alternative to artificial neural networks) was developed ac-
there were probably more than 200 algorithms of this kind tively by the scientific community. The name of this dis-
in books, papers and proceedings. Many researchers dis- cipline, that was coined later, was Evolutionary compu-
cuss that this research area is becoming flooded with new tation. Evolutionary computation was based on the natu-
algorithms that are in fact the old algorithms in a new dis- ral evolution of species and respected the theory of Charles
guise. Potentially, such behavior could be leading into the Darwin. Initially, the Evolutionary Algorithms (EAs) sim-
emergence of pseudoscience. In this paper, we try to find ulated the operators of mutation and crossover, where the
some answers to the questions what lead authors to pro- individuals to survive were selected according to their fit-
pose and develop the new nature-inspired algorithms and ness values. The fitness value was determined according to
what their benefits in doing so are. We propose ways in the evaluation of the fitness function that corresponded to
which to stop the emergence of new algorithms. In line with the problem to be solved. Nevertheless, over the years the
this, we have found that proposing the new population-based EAs were divided into the following kind of algorithms: Ge-
nature-inspired algorithms is actually similar to the swarm netic algorithms [18], evolution strategies [1], genetic pro-
intelligence behavior in nature, where the role of population gramming [16] and evolutionary programming [26]. The
members is acted by authors of the new algorithm with the main differences between these sub-families were basically
goal to publish a paper, thus promoting its algorithm and in the representation of individuals, e.g., binary represen-
spreading it all over the world. tation was used by genetic algorithms, floating point repre-
sentation by evolution strategies, finite state automata by
Keywords evolutionary programming and programs in Lisp by genetic
programming. Additionally, it is important to mention that
metaphors, nature-inspired algorithms, swarm intelligence in the 80s other metaheuristics were also designed [23, 8, 9,
10, 15]. The period when these methods appeared in the
1. INTRODUCTION literature was a little bit calmer compared with nowadays.
It was a time without the Internet and also access to the
Approximately 50 years ago, the time emerged when scien- papers was limited. Additionally, in these times people did
tists began applying algorithms solving the problems on dig- not yet know the term Publish or perish [19, 2]. Scien-
ital computers by mimicking the human brain widely. These tists should not have to be forced to publish for any price
methods were called Artificial neural networks [13]. Ar- in order to their hold position at the university or scientific
tificial neural networks were proposed in the 40s in the pre- institute. But things were changed quickly. The years of 90s
vious century, but it took some time before the community came rapidly. In this scientific area a new paradigm was pro-
began to use them widely for scientific and first practical posed that incorporated the social behavior of many agents
usage. These networks were really interesting methods and that guided them into complex behavior. The roots of this
many scientists claimed that artificial neural networks would method, which is named Swarm intelligence, can be found
power the world in the near future. Artificial neural net- in the dissertation of Marco Dorigo [4]. His method proposed
works were counted into pure artificial intelligence and now the colonies of ants for solving discrete optimization prob-
there are many various types of these networks for solving lems. A little bit later, in 1995, Kennedy and Eberhart [14]
particular tasks in theory and practice. That historical time applied the behavior of bird swarms and fish schools into
was also the time where people were limited with hardware an algorithm with the name Particle swarm optimiza-
tion. These two methods were the beginners of the new
community movement, i.e. the so-called swarm intelligence
community. However, in the 90s and early 2000s the com-
munity did not think that these two powerful algorithms
were the stepping stones for the development of uncount-
able nature-inspired algorithms and, potentially, the flood
of algorithms that led into pseudoscience. In this paper, we
StuCoSReC Proceedings of the 2016 3rd Student Computer Science Research Conference 33
Ljubljana, Slovenia, 12 October
month: Is the current era coming to the end?
Iztok Fister Jr., Uroš Mlakar, Janez Brest, Iztok Fister
Faculty of Electrical Engineering and Computer Science, University of Maribor
Smetanova 17, 2000 Maribor, Slovenia
iztok.fister1@um.si
ABSTRACT and software. Thus, people were unable to test and de-
velop their methods widely. However, some years after the
Every month at least one new population-based nature-insp- creation of artificial neural networks, another discipline (or
ired algorithm has been released in literature. Until recently, alternative to artificial neural networks) was developed ac-
there were probably more than 200 algorithms of this kind tively by the scientific community. The name of this dis-
in books, papers and proceedings. Many researchers dis- cipline, that was coined later, was Evolutionary compu-
cuss that this research area is becoming flooded with new tation. Evolutionary computation was based on the natu-
algorithms that are in fact the old algorithms in a new dis- ral evolution of species and respected the theory of Charles
guise. Potentially, such behavior could be leading into the Darwin. Initially, the Evolutionary Algorithms (EAs) sim-
emergence of pseudoscience. In this paper, we try to find ulated the operators of mutation and crossover, where the
some answers to the questions what lead authors to pro- individuals to survive were selected according to their fit-
pose and develop the new nature-inspired algorithms and ness values. The fitness value was determined according to
what their benefits in doing so are. We propose ways in the evaluation of the fitness function that corresponded to
which to stop the emergence of new algorithms. In line with the problem to be solved. Nevertheless, over the years the
this, we have found that proposing the new population-based EAs were divided into the following kind of algorithms: Ge-
nature-inspired algorithms is actually similar to the swarm netic algorithms [18], evolution strategies [1], genetic pro-
intelligence behavior in nature, where the role of population gramming [16] and evolutionary programming [26]. The
members is acted by authors of the new algorithm with the main differences between these sub-families were basically
goal to publish a paper, thus promoting its algorithm and in the representation of individuals, e.g., binary represen-
spreading it all over the world. tation was used by genetic algorithms, floating point repre-
sentation by evolution strategies, finite state automata by
Keywords evolutionary programming and programs in Lisp by genetic
programming. Additionally, it is important to mention that
metaphors, nature-inspired algorithms, swarm intelligence in the 80s other metaheuristics were also designed [23, 8, 9,
10, 15]. The period when these methods appeared in the
1. INTRODUCTION literature was a little bit calmer compared with nowadays.
It was a time without the Internet and also access to the
Approximately 50 years ago, the time emerged when scien- papers was limited. Additionally, in these times people did
tists began applying algorithms solving the problems on dig- not yet know the term Publish or perish [19, 2]. Scien-
ital computers by mimicking the human brain widely. These tists should not have to be forced to publish for any price
methods were called Artificial neural networks [13]. Ar- in order to their hold position at the university or scientific
tificial neural networks were proposed in the 40s in the pre- institute. But things were changed quickly. The years of 90s
vious century, but it took some time before the community came rapidly. In this scientific area a new paradigm was pro-
began to use them widely for scientific and first practical posed that incorporated the social behavior of many agents
usage. These networks were really interesting methods and that guided them into complex behavior. The roots of this
many scientists claimed that artificial neural networks would method, which is named Swarm intelligence, can be found
power the world in the near future. Artificial neural net- in the dissertation of Marco Dorigo [4]. His method proposed
works were counted into pure artificial intelligence and now the colonies of ants for solving discrete optimization prob-
there are many various types of these networks for solving lems. A little bit later, in 1995, Kennedy and Eberhart [14]
particular tasks in theory and practice. That historical time applied the behavior of bird swarms and fish schools into
was also the time where people were limited with hardware an algorithm with the name Particle swarm optimiza-
tion. These two methods were the beginners of the new
community movement, i.e. the so-called swarm intelligence
community. However, in the 90s and early 2000s the com-
munity did not think that these two powerful algorithms
were the stepping stones for the development of uncount-
able nature-inspired algorithms and, potentially, the flood
of algorithms that led into pseudoscience. In this paper, we
StuCoSReC Proceedings of the 2016 3rd Student Computer Science Research Conference 33
Ljubljana, Slovenia, 12 October
