一个计算积分的替代抽样方法  

An Alternative Algorithm for Estimating Integrals

作　　者：胡果荣[1] 史宁中[2] 张宝学[2] 

机构地区：[1]吉林大学数学研究所,长春130012 [2]东北师范大学数学与统计学院,长春130024

出　　处：《吉林大学学报（理学版）》2006年第3期362-366,共5页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金(批准号:10501005);东北师范大学校内青年基金(批准号:111494074).

摘　　要：为提高高维积分的计算速度,提出一种替换MonteCarlo积分方法.将积分区域以网格的形式离散化,再在网格上以相应的密度函数之值为权函数采用离散的Gibbs抽样算法抽样,对抽样得到的样本作均匀扰动后就可获得所需的新抽样序列,从而得到积分的近似估计值.模拟表明新算法计算速度较快.To enhance the computing speed of high dimensional integral, an alternative Monte Carlo sampling algorithm is proposed in this paper. Firstly, the integral region is partitioned into net form. Secondly, grid points are sampled by using discrete Gibbs sampling method whose weighted functions are values of corresponding density function. Lastly, a new sampling sequence will be obtained by adding an uniform variable sequence to the original sequence correspondingly, and an estimation of the integral is given. The new sampling algorithm is as simple as the traditional numerical method. Simulating output showed that the new algorithm performs very well in computing speed.

关 键 词：模拟 重要性抽样 GIBBS抽样 格子点 数值积分 

分 类 号：O212.4[理学—概率论与数理统计]