基于MPI的精确非参数检验之Fisher网络算法的并行算法研究
Parallel Study of Fisher Network Algorithm for Exact Nonparametric Test Based on MPI
-
摘要: 非参数检验是样本为非正态分布或未知分布情况下进行的假设检验, 适应性强, 稳定性好, 在医学、工程、金融等各种统计场合都有广泛应用.精确非参数检验提高了计算的精度, 但是直接计算精确很费时间.文中针对精确非参数检验主要使用的Fisher网络算法, 基于MPI提出并行算法并予以实现.实验结果表明并行效率良好, 能够大大加快计算的速度.Abstract: Non-parameter test is used for non-normal distribution or unknown distribution samples.It has strong adaptability, good stability, and is widely used in medicine, engineering, finance field etc.Exact non-parameter test improves the calculation precision but takes a lot of time.This article proposes and implements parallel algorithms of exact p-value based on Fisher network algorithm.