信息技术领域中大型稀疏矩阵.pdf

摘 要 在科学和工程计算领域中，有许多问题最终归结为稀疏线性方程组的求解。 在实际应用中，经常需要求解百阶、千阶甚至万阶及以上的稀疏线性方程组，完 成大量的矩阵数值运算，这是数值计算研究的热点问题。本文主要研究三个方面 的内容：一是分析比较大型稀疏矩阵的几种压缩存储方法：CSR 方法，CCS 方法， JSA 方法，Shermans 方法，Ellpack-Itpack 方法，以及它们对空间的利用率， 并对有关方法作了相应改进，减少存储量，提高运算速度；二是针对大型稀疏对 称矩阵，给出合同对角化方法，该方法便于计算机编程实现；三是分别利用迭代 -Cholesky 混合法和迭代-追赶混合法解决对流扩散中的稀疏线性方程组问题， 并给出相关的算法程序，结合算例，表明该方法有较好的数值效果。 关键词：稀疏矩阵，压缩存储，Cholesky 分解，迭代法，对称矩阵，合同变换 Theory and Solution of Large and Sparse Matrix Equations in IT Field Abstract In the computational field of science and engineering, many problems finally go to solve the sparse linear equations that has hundreds of thousands of ranks and complete a lot of matrix algebra operations. It’s a popular subject in algebra operations nowadays. This paper consists of three parts. First is to analyze and compare several compressed storage methods that are the CSR/CCS method, the JSA method, the Sherman ’s method, the Ellpack-Itpack method, and their space utilization, and streamline some relevant methods to decrease the storage space and increase the computer speed. The second is to provide the method of congruent diagonalization for large sparse symmetric matrix, it can be realized by computing program. The third is to apply two methods to deal with the sparse linear equations in the convection-diffusion respectively, one is to combine the Cholesky factorization with the iterative method, the other is to combine Pursuit factorization with the iterative method, and provid e the corresponding program. With the numerical examples, it shows the methods have good algebra results. Key words ：Sparse matrix, Compressed storage, Cholesky factorization, Iterative method, Sy