EFFICIENT ALGORITHMS FOR STATISTICAL CALCULATION IN HETEROGENIC PARALLEL DISTRIBUTED COMPUTATIONAL ENVIRONMEN

点击翻译按钮获取中文摘要

Possibilities of parallelization for algorithms in statistical calculation are explored in a parallel distributed computational environment where various PCs in a laboratory are inter-connected via LAN. Efficiency and remedies of pre-distribution and work-pool approaches are examined as a basis of task distribution scheme in the context of Monte-Carlo integration. Then, appropriateness of their parallelism approaches is evaluated for calculation algorithms for statistical methods, such as, exact and Bayes inferences for logistic regression analysis and classification tree based approach.