What is the best method to solve large linear system equation (Ax=b) (A:million~ by million~)

12 次查看（过去 30 天）

显示更早的评论

Youngsuk Jung 2018-12-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/433943-what-is-the-best-method-to-solve-large-linear-system-equation-ax-b-a-million-by-million

评论： Christine Tobler 2018-12-6

采纳的回答： Christine Tobler

The maxtix A is sparse from finite element analysis.

I have used backslash to solve the equation but I want to reduce the calculation time.

I also tried to use GPU acceleration but there is a memory limitation with Titan xp with 12Gb memory.

采纳的回答

Christine Tobler 2018-12-5

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/433943-what-is-the-best-method-to-solve-large-linear-system-equation-ax-b-a-million-by-million#answer_350727

You can try one of the Iterative Methods and Preconditioners.

3 个评论
显示 1更早的评论隐藏 1更早的评论

Bruno Luong 2018-12-6

Why? MATLAB has also a bunch of iterative solvers to chose from: cgs, bicg, gmres, etc...

Christine Tobler 2018-12-6

MATLAB chooses an appropriate direct solver in backslash. Using iterative solvers requires more trial and error, and more knowledge of the original system (as preconditioners are often necessary). For a large, well-conditioned matrix, there can be a significant speed-up compared to backslash, if a suitable iterative solver is used.

请先登录，再进行评论。