摘要
针对应用学科仿真实验中所需高斯随机数质量要求日趋严格的问题,对比研究了高斯随机数常用的中心极限生成算法、Box-Muller算法和极化判决算法,研究了各算法在尾部区域内生成高斯随机数的质量.仿真结果表明:极化判决算法为最佳选择,在增加一定量运算代价下,换取高斯随机数的高尾部精度.需注意极化判决算法中由于用到判决语句,导致高斯随机数的产生速度不恒定,硬件实现时需用先入先出缓冲器解决该问题.
Aiming at the quality of Gaussian random numbers in simulation experiments for various applied sciences,comparative study was made on the generate algorithms used to generating Gaussian random number, focusing on the quality of each algorithm to generate Gaussian random numbers in the tail region.Simulation results show that,namely to increase the operational costs,Polar-Rejection algorithm may well be the best option with high- tail accuracy.It should be noted that,due to using if-else statements in Polar-Rejection algorithm,resulting the rate of Gaussian random numbers is not constant.It should be using first-in first-out buffer to solve the problem in hardware implementation.
出处
《河南科技学院学报（自然科学版）》
2014年第3期56-59,共4页
Journal of Henan Institute of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(U1204606)
河南省基础与前沿技术研究计划项目(142300410004)
河南省教育厅科学技术研究重点研究项目(14B510020)
关键词
高斯随机数
模拟
Ganssian random number
simulation
