Re: [eigen] qr().solve() implementation

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


On Wed, Jan 28, 2009 at 1:48 AM, Gael Guennebaud
<gael.guennebaud@xxxxxxxxx> wrote:
> Hi,
>
> thanks for the patch, however there were some shortcomings because you
> assumed the input matrix was full rank.

Yes, I realized that the patch was incomplete; I sent it more for
comment than to be committed as is. Thanks for adding the
invertability check.

Note that the file doc/snippets/QR_solve.cpp doesn't appear to have
been committed, so I don't believe that part of the docs is correct.

Keir

> here is the commit log:
>
> [10:46] <CIA-60> ggael * r917639 eigen2/trunk/kdesupport/eigen2/ (4
> files in 3 dirs):
> [10:46] <CIA-60> QR: add isInjective(), isSurjective(),
> [10:46] <CIA-60> mark isFullRank() deprecated,
> [10:46] <CIA-60> add solve() (mix of Keir's patch and LU::solve())
> [10:46] <CIA-60> => there is big problem with complex which are not working
>
> Gael
>
> On Tue, Jan 27, 2009 at 7:33 PM, Keir Mierle <mierle@xxxxxxxxx> wrote:
>> Here is a (incomplete) implementation of qr().solve(b, &x).
>>
>> There is something wrong with the generated docs; it seems to take the
>> results from LU_solve.cpp rather than QR_solve.cpp but I don't have
>> time to figure this out now.
>>
>> It is probably wasteful in that it explicitly computes matrixQ() and
>> matrixR(). However, it does work.
>>
>> Keir
>>
>
>
>



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