三种混合二维正态分布代表点生成方法的比较

Comparison of three generation methods of representative points of 2D mixed normal distribution

本文对三种寻找混合二维正态分布代表点的算法:蒙特卡洛法、数论法及均方误差法所产生的近似总体与真实总体间的分布偏差及均方误差进行了比较研究,结果表明均方误差法是最优的.然后本文采用自助法对三类代表点进行重抽样,并以重抽样的结果再次验证了均方误差法的优越性.

In this paper, three generation methods representative points （RPs） of 2D mixed normal dis- tribution are given. These methods are based on the Monte Carlo method, the number-theoretical meth- od and the mean square error （MSE） method respectively. Then, performance of the methods are com- pared in sense of distribution discrepancy and mean square error criterion. It is shown that the MSE method is the best one. Moreover, a comparison with the bootstrap technique, which is used to resam- ple the RPs, also shows superiority of the MSE method.