The Block Jacobi Method for Computing the Singular Value Decomposition
The Block Jacobi Method for Computing the Singular Value Decomposition
复制标题
计算奇异值分解的分块雅可比法
DOI：
--
复制
发表时间：
1985
期刊：
影响因子：
--
通讯作者：
C. Loan
中科院分区：
文献类型：
--
作者：
C. Loan
用于计算对称特征值和奇异值分解的雅各比技术由于对并行计算的兴趣而实现了最近的突出。它们非常适合某些具有以最近的邻居方式连接的处理器的多处理器系统。如果处理器相当强大并且具有很大的本地内存，则阻止雅各比程序很有吸引力，因为它们会导致更有利的计算与通信比率。本文研究了与两个块雅各比方法相关的一些实用细节，以进行单数值分解。这些方法在2 BB-2子问题如何求解方面有所不同。
Jacobi techniques for computing the symmetric eigenvalue and singular value decompositions have achieved recent prominence because of interest in parallel computation. They are ideally suited for certain multiprocessor systems having processors that are connected in nearest neighbor fashion. If the processors are reasonably powerful and have significant local memory, then block Jacobi procedures are attractive because they render a more favorable computation to communication ratio. This paper examines some of the practical details associated with two block Jacobi methods for the singular value decomposition. The methods differ in how the 2-by-2 subproblems are solved.