广西大学学报(自然科学版)
廣西大學學報(自然科學版)
엄서대학학보(자연과학판)
JOURNAL OF GUANGXI UNIVERSITY (NATURAL SCIENCE EDITION)
2014年
2期
341-345
,共5页
K-Medoids算法%并行算法%计算粒度%MapReduce
K-Medoids算法%併行算法%計算粒度%MapReduce
K-Medoids산법%병행산법%계산립도%MapReduce
K-Medoids algorithm%parallel algorithm%computing grain%MapReduce
K-Medoids算法具有不同层次的并行性,计算粒度不同对并行算法效率有较大影响。基于K-Me doids的并行计算特点,提出了一个改进的K-Medoids并行算法,该算法基于MapReduce模型,通过适当增加计算粒度,降低了通信消耗占比。实验结果表明,改进的并行算法与其他已有算法相比,加速比与运行效率有显著提高。
K-Medoids算法具有不同層次的併行性,計算粒度不同對併行算法效率有較大影響。基于K-Me doids的併行計算特點,提齣瞭一箇改進的K-Medoids併行算法,該算法基于MapReduce模型,通過適噹增加計算粒度,降低瞭通信消耗佔比。實驗結果錶明,改進的併行算法與其他已有算法相比,加速比與運行效率有顯著提高。
K-Medoids산법구유불동층차적병행성,계산립도불동대병행산법효솔유교대영향。기우K-Me doids적병행계산특점,제출료일개개진적K-Medoids병행산법,해산법기우MapReduce모형,통과괄당증가계산립도,강저료통신소모점비。실험결과표명,개진적병행산법여기타이유산법상비,가속비여운행효솔유현저제고。
K-Me doids algorithm has different levels of parallelism, and the computing grain has im-portant impact on the efficiency of the parallel algorithm. With the consideration of the parallelism features, an improved parallel K-Medoids algorithm is presented. Based on the MapReduce model, the improved algorithm has bigger computing grain and reduces the percentage of communication cost. The experimental results show that the new algorithm obtains better speedup and efficiency comparing with other algorithms.