期刊名称：TELKOMNIKA (Telecommunication Computing Electronics and Control)
印刷版ISSN：2302-9293
出版年度：2016
卷号：14
期号：3
页码：1099-1104
DOI：10.12928/telkomnika.v14i3.3609
语种：English
出版社：Universitas Ahmad Dahlan
摘要：Artificial Bee Colony(ABC) or its improved algorithms used in solving high dimensional complex function optimization issues has some disadvantages, such as lower convergence, lower solution precision, lots of control parameters of improved algorithms, easy to fall into a local optimum solution. In this letter, we propose an improved ABC of staged search. This new algorithm designs staged employed bee search strategy which makes that employed bee has different search characters in different stages. That reduces probability of falling into local extreme value. It defines the escape radius which can guide precocious individual to jump local extreme value and avoid the blindness of flight behavior. Meanwhile, we adopt initialization strategy combining uniform distribution and backward learning to prompt initial solution with uniform distribution and better quality. Finally, we make simulation experiments for eight typical high dimensional complex functions. Results show that the improved algorithm has a higher solution precision and faster convergence rate which is more suitable for solving high dimensional complex functions.
其他摘要：Artificial Bee Colony(ABC) or its improved algorithms used in solving high dimensional complex function optimization issues has some disadvantages, such as lower convergence, lower solution precision, lots of control parameters of improved algorithms, easy to fall into a local optimum solution. In this letter, we propose an improved ABC of staged search. This new algorithm designs staged employed bee search strategy which makes that employed bee has different search characters in different stages. That reduces probability of falling into local extreme value. It defines the escape radius which can guide precocious individual to jump local extreme value and avoid the blindness of flight behavior. Meanwhile, we adopt initialization strategy combining uniform distribution and backward learning to prompt initial solution with uniform distribution and better quality. Finally, we make simulation experiments for eight typical high dimensional complex functions. Results show that the improved algorithm has a higher solution precision and faster convergence rate which is more suitable for solving high dimensional complex functions.
