inverse of large size sparse matrix
in [Matlab]
|
From: Zuofeng on 24 Dec 2009 22:47 Hi Tim, I have a question about the inverse of a large dimensional square matrix. I have a matrix of 2048 x 2048, and I need to invert it in each of 5000 iterations, which is time consuming. Basically, I need to calculate A * inv(B)* A' in each loop. B is a large matrix. Is there any fast way to do this? Please forgive me if this question is kind of stupid and accept my best thanks! Jeff
From: Zuofeng on 24 Dec 2009 22:47 Hi Tim, I have a question about the inverse of a large dimensional square matrix. I have a matrix of 2048 x 2048, and I need to invert it in each of 5000 iterations, which is time consuming. Basically, I need to calculate A * inv(B)* A' in each loop. B is a large matrix. Is there any fast way to do this? Please forgive me if this question is kind of stupid and accept my best thanks! Jeff
From: John D'Errico on 25 Dec 2009 06:04 "Zuofeng " <shang(a)stat.wisc.edu> wrote in message <hh1cjp$n8e$1(a)fred.mathworks.com>... > Hi Tim, > > I have a question about the inverse of a large dimensional square matrix. > I have a matrix of 2048 x 2048, and I need to invert it in each of 5000 iterations, > which is time consuming. > > Basically, I need to calculate A * inv(B)* A' in each loop. B is a large matrix. > Is there any fast way to do this? Please forgive me if this question is kind of stupid > and accept my best thanks! > > Jeff Do you have a fixed matrix B that you are repeatedly trying to invert here? If so, then don't invert it. Instead, factorize it. (You can use Tim's factorize utility from the file exchange to do so.) http://www.mathworks.com/matlabcentral/fileexchange/24119 Store the matrix factorization, then just re-use it repeatedly. If B is changing inside the loop, then you can still use a form like A*(B\A') instead of using inv. John
|
Pages: 1 Prev: information lost while showing image on axes Next: Decision Tree Modeling in Matlab |