Re: [eigen] Using SparseLU to calculate the determinant.

[ Thread Index | Date Index | More Archives ]

A simple approach is to set a flag before the analyze/factorize phase to let the solver compute the left and right diagonal matrices Dr and Dc matrices. Something like
SparseLU<MatrixType> lu;
lu.setEquil(boolean); // set the flag to compute or not the scaling matrices lu.compute (A); // Compute PrDrADcPc = LU where Pr and Pc are the row and column permutations.

On 04/15/2013 07:25 PM, Gael Guennebaud wrote:
Do you have an idea to add balancing to SparseLU (API-wise)? We should
make sure whether this can be easily added without breaking the API or
requiring a new class.


On Mon, Apr 15, 2013 at 5:52 PM, Desire NUENTSA
<desire.nuentsa_wakam@xxxxxxxx>  wrote:
Please, find a feature request at
with a patch to do that.
In my opinion, We should add a feature to balance the matrix before pushing
the patch.

Nuentsa W.

On 04/13/2013 05:39 PM, Joseph Maestri wrote:


Decided to email this since SparseLU is in 3.2.0-beta1.

Currently, calculating the determinant of a sparse matrix can be achieved
using the SparseCholesky module. Is there a plan to extend this
functionality to the SparseLU module? Not sure how trivial this is since I
have not looked into the specifics of the implementation and how the
factorisation is stored.


Mail converted by MHonArc 2.6.19+