单种子和多种子对正态随机序列质量的影响  

Influence of single seed and multi-seed on quality of a normal distribution random sequence

作　　者：赵学智[1] 叶邦彦[1] 

机构地区：[1]华南理工大学机械与汽车工程学院,广州510640

出　　处：《振动与冲击》2011年第1期49-55,共7页Journal of Vibration and Shock

基　　金：国家自然科学基金资助项目(50875086);广州市科技计划资助项目(2008J1-C101)

摘　　要：通常一个正态随机序列是由同一个种子生成的,然而实际上序列中的每一个随机数还可用不同的种子来产生。提出了多种子的产生方法以及连续调用随机数发生器时防止产生相同序列的措施,利用统计模拟和假设检验理论,采用四个不同的统计量从正态性、均值、方差、随机性这四个方面对单、多种子法在不同时间段生成的大量随机序列的质量进行了对比研究,结果表明,多种子法生成的随机序列在正态性和方差质量方面有明显的优势,而在均值和随机性方面,这两种方法的效果基本一致,故总的说来多种子法要优于单种子法。Generally a normal distribution random sequence is generated with a single seed, but in fact, each random number in this sequence can also be generated with different seeds. The method to generate different seeds was put forward and the measure to avoid generating same sequences when the random number generator was used continuously was proposed. Four statistical parameters were used, the qualities of a great lot of normal random sequences generated with single seed and multi-seed methods respectively at different time were contrastively studied in the aspects of normality, mean, deviation and randomness by virtue of statistical simulation and hypothesis test theory. The results showed that random sequences generated with multi-seed method have the evident advantage in normality and deviation quality, while in mean and randomness quality, the effects of these two methods are basically same; so as a whole, the multi-seed method is better than the single seed one.

关 键 词：正态随机序列 单种子 多种子 假设检验 统计模拟 

分 类 号：TN911.7[电子电信—通信与信息系统] TH165.3[电子电信—信息与通信工程]