Computing generalized Inverse of a square but sparse matrix?

I am trying to find the generalized inverse of a square and sparse matrix. I am using Matlab, however pinv(matrix) doesn't work for sparse matrices. If you have any suggestion or algorithm, please share. Thanks in advance.

Answer 1

[Rishabh Mishra]

Hi,

The disadvantages of using ‘pinv’ to find inverse of sparse matrix are:

1. ‘pinv’ requires costly SVD.
2. ‘pinv’ does not work with sparse matrix.

Instead, you can use the function ‘pseudo-inverse’ for the same purpose. Refer the following link for documentation of ‘pseudo-inverse’ function.

Hope this helps!

Answer 2

[Christine Tobler]

If you want to apply pinv(A)*b, you can instead use lsqminnorm(A, b), which also works for sparse matrices and does something equivalent (using QR decomposition instead of SVD).