Re: [eigen] [patch] LDLt decomposition with rank-deficient matrices

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]


2010/2/23 Ben Goodrich <bgokgm@xxxxxxxxxxxxxx>:
> Hi Benoit,
>
> On Tue, Feb 23, 2010 at 4:11 PM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
>
>> Both bugs fixed in the development branch.
>
> Thank you for fixing this. I agree that the previous test case is
> working now. You might have a look at the stable branch to see if it
> has the behavior you want, although it is assuming positive
> definiteness anyway. I would check myself except that this backported
> fix for the original problem of ldlt() not accepting Maps
>
> https://bitbucket.org/eigen/eigen/changeset/901a65334944/
>
> makes the stable branch not compile. I get these errors compiling that
> test case:
>
> In file included from /tmp/eigen/Eigen/Cholesky:36,
>                 from test.cc:1:
> /tmp/eigen/Eigen/src/Cholesky/LDLT.h: In member function ‘const
> Eigen::LDLT<typename Eigen::MatrixBase<Derived>::PlainMatrixType>
> Eigen::MatrixBase<Derived>::ldlt() const’:
> /tmp/eigen/Eigen/src/Cholesky/LDLT.h:195: error: ‘PlainObject’ was not
> declared in this scope
> /tmp/eigen/Eigen/src/Cholesky/LDLT.h:195: error: template argument 1 is invalid
>
> This just looks like an #include thing, but I will let someone more
> experienced with eigen patch it.

Fixed, thanks for the heads up.

Also, I committed a second precision fix to the 2.0 branch. It should
now be really precise. The trade-off is that isPositiveDefinite() now
always returns true (it never worked anyway, since this decomposition
is not rank-revealing, especially as in 2.0 it doesn't even use any
pivoting).

Benoit

>
> Thanks,
> Ben
>
>
>



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/