An Efficient Artificial Bee Colony Algorithm with an Improved Linkage Identification Method
Hao Gao,Zheng Fu,Chi-Man Pun,Jun Zhang,Sam Kwong
DOI: https://doi.org/10.1109/tcyb.2020.3026716
IF: 11.8
2020-01-01
IEEE Transactions on Cybernetics
Abstract:The artificial colony (ABC) algorithm shows a relatively powerful exploration search capability but is constrained by the curse of dimensionality, especially on nonseparable functions, where its convergence speed slows dramatically. In this article, based on an analysis of the difference between updating mechanisms that include both all-variable and one-variable updating mechanisms, we find that when equipped with the former strategy, the algorithm rapidly converges to an optimal region, while with the latter strategy, it searches the solution space thoroughly. To utilize multivariable and one-variable updating mechanisms on nonseparable and separable functions, respectively, we embed an improved linkage identification strategy into the ABC by detecting the linkage between variables more effectively. Then, we propose three common strategies for ABC to improve its performance. First, a new approach that considers the historic experiences of the population is proposed to balance exploration and exploitation. Second, a new strategy for initializing scout bees is used to reduce the number of function evaluations. Finally, the individual with the worst performance is updated with a defined probability on multiple dimensions instead of one dimension, causing it to follow the population steps on nonseparable functions. This article is the first to propose all these concepts, which could be adopted for other ABC variants. The effectiveness of our algorithm is validated through basic, CEC2010, CEC2013, and CEC2014 functions and real-world problems.