J4

• 数学 • 上一篇    下一篇

求解抛物型方程的并行Monte Carlo区域分解算法

刘芳芳1,2, 刘播2   

  1. 1. 中国科学院 软件研究所, 北京 100083; 2. 吉林大学 数学研究所, 长春 130012
  • 收稿日期:2006-06-26 修回日期:1900-01-01 出版日期:2007-03-26 发布日期:2007-03-26
  • 通讯作者: 刘芳芳

Parallel Monte Carlo Domain Decomposition Algorithm for Solving Parabolic Functions

LIU Fangfang1,2, LIU Bo2   

  1. 1. Institute of Software, Chinese Academy of Sciences, Beijing 100083, China;2. Institute of Mathematics, Jilin University, Changchun 130012, China
  • Received:2006-06-26 Revised:1900-01-01 Online:2007-03-26 Published:2007-03-26
  • Contact: LIU Fangfang

摘要: 介绍了用Monte Carlo方法求解抛物型方程的3种游动模型, 给出了相应的证明及误差的概率估计式; 将Monte Carlo方法和区域分解算法相结合提出一种可并行计算抛物型方程的方法, 针对形式一般的方程给出了具体算法, 并指出算法适用的条件; 分别对二维、 三维抛物型方程进行数值实验, 实验结果表明该算法通过合理的安排, 几乎不需要数据传递, 在并行机上可以节省大量的计算时间.

关键词: MonteCarlo方法, 区域分解, 并行计算, 抛物型方程

Abstract: This paper introduces three random walk models on computating parabolic equations solved by Monte Carlo method, the corresponding prove and the error estimation are given; then a method on computating parabolic functions parallel is proposed based on the combination of Monte Carlo method with domain decomposition algorithm, the algorithm for onedimension functions is expounded. The condition suitable to the algorithm is given. Some numerical experiments were carried out for 2Dand 3Dparabolic functions. The resultsshow that the method can well handle parabolic functions parallel and can save lots of time.

Key words: Monte Carlo method, domain decomposition, parallel computation, parabolic equation

中图分类号: 

  • O242.1