一种基于密度的并行聚类算法

Efficient parallel clustering algorithm based on density

针对微阵列基因表达数据聚类的高维复杂性,提出了一种基于密度的并行聚类算法,在APRAM模型的分布式存储系统中,通过欧几里德距离矩阵和密度函数两次时间复杂度为O(np2)的计算,可使聚类过程的时间复杂度为O(npK),以增加一次计算的代价来降低聚类过程的时间复杂度。基于8结点的机群计算实验表明:本算法能够达到较同类算法更高的并行加速比,提高高维生物数据的聚类速度。

Aim at the high complexity of the gene expression data clustering,puts forward a parallel clustering algorithms based on the density.Uses MPI under the APRAM model,passing two compute with parallel time complexity is O（n^2/P ） that of the P Euclidean distance matrix and the density function,can make the parallel time complexity of clustering be O（nK/P）,reduces the P time complexity of clustering through adding one compute.The experiment based on eight nodes indicates that this algorithm can attain higher parallel accelerate ratio than the same kind algorithm,raise the clustering rate of the high dimension living data.