
Re: Solving sparse Ax=b without certain eigenvector spaces
Posted:
Aug 2, 2014 8:37 PM


"Hui " <huiprobable@gmail.com> wrote in message <lrjnpr$klo$1@newscl01ah.mathworks.com>... > Hi, I'm trying to solve a sparse Ax=b problem. The size of A is about 10^4 by 10^4. A is pretty singular and the first a few eigenvalues of A is about 10^(12), 10^(6) and 10^(4). I'm able to find the first a few eigenvalues and eigenvectors using the command eigs, which takes only one or two seconds. Now I want to find x, such that Ax=b, but I want to exclude the first three eigenvectors. > > Are there any simple ways that allow me to do this?
Pretty singular? Excluding eigenvectors? Please define your terms, as pretty singular does not have a clear definition.
I'm also perplexed by your statement that the first eigenvalue you list is the smallest of the bunch, but also what your goals are here.
My guess is you wish to deflate the matrix, removing those first few eigencomponents from A.
john

