当前Spark分布式编程框架由于内存计算得到了快速发展，相对于传统MapReduce并行编程模型在迭代运算上有明显优势。针对串行遗传算法处理大规模问题能力有限的现状，提出了一种基于Spark平台的粗粒度并行遗传算法（SPGA）。该方法利用Spark框架并行实现了遗传算法的选择、交叉和变异操作，并对并行操作算子的性能进行了分析，优化了算法并行化实现方案，极大地提高了遗传算法全局搜索效率。实验结果表明，新的并行遗传算法在收敛速度上有显著的提高，能够很好地提高优化效率。%Spark parallel programming framework has been developed rapidly because of the memory computing, compared with traditional parallel programming based on MapReduce, it has obvious advantages in the speed of iteration. This paper proposes a coarse-grained parallel genetic algorithm based on Spark to solve the problem that serial genetic algorithm encounters. In this paper, the algorithm is used to realize the selection, crossover and mutation operation of genetic algorithm by parallel programming operators, the performance of the parallel operator is analyzed and the parallel design scheme of the algorithm is optimized, which greatly improves the search efficiency of genetic algorithm. Experimental results show that the new parallel genetic algorithm can improve convergence rate and economize calculating time.
展开▼
