- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I have gone back to the developers of sparse solver and they would like to understand why you want this particular quantity. Clearly it would be possible to add such functionality, but if they can understand the unlying motivation for this they may be able to address that.
Bruce
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
There are many situations where the factors of the Cholesky decompositon are needed. One of them, very important for many number smashers, regards the solution of the generalized eigenvalue problem:
find (L, X) (L real and X vector) such that
K X = L M X
where K,M are symmetric positive definite matrices.
The generalized problem could be transformed into a classical eigenvalue problem, retaining the symmetry and posiveness of the matrix operator A:
A X = 1/L X
where A depends on K,M. Using A Lanczos or a subspace iteration method for this problem (when coming from the generalized one) requires access to the factors of K (or M).
Please note that generalized eigenvalue analysis is a very common task in many engineering sciences.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
This is exactly the information I need. I can't say for certain when I will have more information on this butI will forward this to the developers.
Bruce
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
has there been any development re. sparse cholesky ?
Thank you.
mepa

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page