Towards the universal framework of stochastic nature-inspired population-based algorithms
Iztok Fister, Janez Brest, Uroš Mlakar · 2016
Stochastic nature-inspired population-based algorithms have attracted a huge community of researchers and practioners for who use them for optimization purposes. The implications of using the methods are almost unlimited. Most of these nature-inspired algorithms are inspired by the biological principles of behavior of various animals living in nature. In our opinion, a lot of research has turned in the wrong direction recently. Instead of improving existing algorithms, they have been proposing the new algorithms discontinuously. Unfortunately, most of these so called “new nature-inspired algorithms” are actually modifications of already known algorithms hidden behind the pompous metaphor-based inspiration. In this theoretically oriented paper, we propose a universal framework of stochastic nature-inspired population-based algorithms that try to unify the more real-coded evolutionary and swarm intelligence based algorithms under the same umbrella. In line with this, a domain-specific embedded language is proposed that enables a generation of the more important stochastic real-coded nature-inspired population-based algorithms on the one hand, and creation of the new ones on the other hand. Consequently, this approach proves that the gap between the nature-inspired algorithms is actually not so big as seems at the first sight.