基于修正的对称分解的稀疏近似逆预处理器

Sparse approximate inverse preconditioners based on a revised Cholesky factorization

摘要: 提出一种基于修正的对称分解(Cholesky)的稀疏近似逆(SAI)预处理技术。对传统的Cholesky分解进行修正,使之能应用于离散电场积分方程所得的复数对称矩阵,然后,用此修正的Cholesky分解为多层快速多极子算法构造SAI预处理器。数值实验表明:基于修正的Cholesky分解的SAI预处理器比基于QR分解的SAI预处理器更高效。

Abstract: Sparse approximate inverse(SAI)preconditioner based on a revised Cholesky factorization is presented in this paper.The traditional Cholesky factorization is firstly revised to cope with the matrix arising from electric field integral equations,which is a complex symmetric matrix,then this revised Cholesky factorization is applied to construct SAI preconditioner for the multilevel fast multipole algorithm(MLFMA).Numerical experiments show that SAI preconditioner constructed by the revised Cholesky factorization performs more efficiently than the previous SAI constructed from QR factorization.