Search

Over the past few decades, there has been a surge of interest of using swarm intelligence (SI) in computer-aided optimization. SI algorithms have demonstrated their efficacy in solving various types of real-world optimization problems. However, it is impossible to find an optimization algorithm that can obtain the global optimum for every optimization problem. Therefore, researchers extensively try to improve methods of solving complex optimization problems. Many SI search algorithms are widely applied to solve such problems. ABC is one of the most popular algorithms in solving different kinds of optimization problems. However, it has a weak local search performance where the equation of solution search in ABC performs good exploration, but poor exploitation. Besides, it has a fast convergence and can therefore be trapped in the local optima for some complex multimodal problems. In order to address such issues, this paper proposes a novel hybrid ABC with outstanding local search algorithm called β-hill climbing (βHC) and denoted by ABC–βHC. The aim is to improve the exploitation mechanism of the standard ABC. The proposed algorithm was experimentally tested with parameters tuning process and validated using selected benchmark functions with different characteristics, and it was also evaluated and compared with well-known state-of-the-art algorithms. The evaluation process was investigated using different common measurement metrics. The result showed that the proposed ABC–βHC had faster convergence in most benchmark functions and outperformed eight algorithms including the original ABC in terms of all the selected measurement metrics. For more validation, Wilcoxon's rank sum statistical test was conducted, and the p values were found to be mostly less than 0.05, which demonstrates that the superiority of the proposed ABC–βHC is statistically significant. [ABSTRACT FROM AUTHOR]

The paper introduces a new approach to enhance optimization algorithms when solving the piecewise linearization problem of a given function. Eight swarm intelligence algorithms were selected to be experimentally compared. The problem is represented by the calculation of the distance between the original function and the estimation from the piecewise linear function. Here, the piecewise linearization of 2D functions is studied. Each of the employed swarm intelligence algorithms is enhanced by a newly proposed automatic detection of the number of piecewise linear parts that determine the discretization points to calculate the distance between the original and piecewise linear function. The original algorithms and their enhanced variants are compared on several examples of piecewise linearization problems. The results show that the enhanced approach performs sufficiently better when it creates a very promising approximation of functions. Moreover, the degree of precision is slightly decreased by the focus on the speed of the optimization process. [ABSTRACT FROM AUTHOR]