随机分形搜索算法-《计算机技术与发展》

文章信息/Info

Author(s):: GE Qian-xing; MA Liang; LIU Yong; University of Shanghai for Science and Technology,Shanghai 200093,China

Keywords:: random fractal; stochastic fractal search; diffusion; update; optimization

摘要:: 现有的元启发式算法大多是模仿生物的群体运动来解决优化问题。为了进一步给优化算法的设计提供新的思路,受自然生长现象的启发,提出了一种新型的元启发式算法—随机分形搜索算法。该算法利用分形的扩散特性进行寻优,其优化原理完全不同于现有的元启发式算法。其中,算法的扩散过程采用高斯随机游走方式来开发问题的搜索空间,而更新过程则分别对个体的分量及个体本身采用相应的更新策略来进行更新,以此进行全局搜索和局部搜索,从而形成了一个完整的优化系统。通过对一系列典型的测试函数优化问题的求解实验并与其他算法进行比较,结果表明随机分形搜索算法不仅具有较高的计算精度,而且具有较快的收敛速度。

Abstract:: Most of the existing meta-heuristics are to imitate the biological group movement to solve the optimization problem. In order to provide a new idea for the design of the optimization algorithm,we propose a new meta-heuristic algorithm,stochastic fractal search (SFS),which is inspired by the natural growth phenomenon. The algorithm uses fractal diffusion to find the optimal,and its optimization principle is quite different from the existing meta-heuristic algorithms. In this algorithm,the diffusion processing adopts the Gauss random walk to exploit the search space of the problem,and the update processing separately updates the individual component and the individual itself with the corresponding update strategy to perform global search and local search,thus forming a complete optimization system. Series of computational experiments on typical benchmark functions are tested and the comparisons with that of other algorithms show that the SFS has both high computational accuracy and faster convergence rate.