基于Spark的并行化头脑风暴优化算法及复杂多峰函数优化

A Spark-based parallel brainstorm optimization algorithm and its application in optimizing complex multimodal functions

头脑风暴优化BSO算法是一种新型的群体智能优化算法,启发于众人集思广益求解问题的模式,适合求解复杂多峰函数优化问题。但是,BSO求解多峰极值时需进行重复的迭代运算,面对大规模数据集时会出现计算效率与求解精度过低的现象。为解决上述问题,设计并实现了一种基于Spark的并行化头脑风暴优化算法,通过将BSO算法中计算复杂度最高的聚类与新解产生过程并行化,以提高算法的加速比与计算效率。特别地,基于并行化思想,将种群划分为多个子群进行协同演化,每个子群独立产生新解来保持种群多样性,提高算法的收敛速度。最后,利用并行化BSO算法求解多峰函数。实验表明,在并行节点的总核心数为10的情况下,并行化BSO算法计算时间节省一半,计算精度和串行BSO算法基本持平,收敛速度明显提高,实验结果说明了并行化BSO的有效性。

The brain storm optimization (BSO) algorithm is a new type of swarm intelligence optimization algorithm, which is inspired by the brainstorm problem-solving model and is suitable for solving complex multimodal function optimization problems. However, when solving multimodal extremum, the BSO requires iterative operations. When computing large data sets, its computation efficiency and accuracy are too low. In order to solve the above problems, we design and implement a parallel brainstorm optimization algorithm based on Spark. By parallelizing the clustering with the highest computational complexity in the BSO algorithm and the generation process of new populations, the speedup and efficiency of the algorithm are improved. In particular, based on the idea of parallelization, the population is divided into multiple subgroups for co evolution, and each subgroup produces new populations to maintain the diversity of the population and improve the convergence speed of the algorithm. Finally, the parallel BSO algorithm is used to solve the multimodal function. Experiments show that, when the total number of cores of the parallel nodes is 10, the computation time of the parallel BSO algorithm is saved by 50%, the computational accuracy is basically equal to the serial BSO algorithm, and the convergence speed is improved obviously. The results prove the validity of the parallelization of the BSO.