An error of matrix inversion using NumPy
Travis Oliphant
oliphant.travis at ieee.org
Wed Apr 4 14:08:18 EDT 2007
lancered wrote:
>
>
> So, can you tell me what goes wrong? Is this a bug in
> Numpy.linalg? How to deal with this situation? If you need, I can
> post the matrix I used below, but it is so long,so not at the moment.
>
As you discovered, it is very likely your problem is a very high
condition number.
The easiest thing to do is to use
numpy.linalg.pinv
to perform a pseudo-inverse which will only use the singular-values that
are well-conditioned to compute the inverse.
This will still not give you an exact identity, but at least you will
know you aren't amplifiying low-valued singular vectors.
-Travis
More information about the Python-list
mailing list