You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@commons.apache.org by Dimitri Pourbaix <po...@astro.ulb.ac.be> on 2010/03/15 16:12:56 UTC
[math] Status of MATH-327?
Hi,
What is the present status of MATH-327? Phil updated on Saturday, I guess,
to say it will be fixed in 2.2. However, the present version adresses the
problem of the number of iterations reaching the maximum. It also improves
(but does not completely solve) the situation mentioned by Luc in January
28: there is still one spurious non zero singular value
(3.547702387229884E-7 which I am going to try to kick out) instead of two.
Dim.
----------------------------------------------------------------------------
Dimitri Pourbaix *
Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
CP 226, office 2.N4.211, building NO * and CARPE DIEM.
Universite Libre de Bruxelles *
Boulevard du Triomphe * Tel : +32-2-650.35.71
B-1050 Bruxelles * Fax : +32-2-650.42.26
http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Luc Maisonobe <Lu...@free.fr>.
Phil Steitz a écrit :
> Dimitri Pourbaix wrote:
>> Phil,
>>
>>>> Right now, I see no option but avoiding the computation of A^tA. It is
>>>> fair to ask for 2.2.
>>> OK, so you are OK with keeping the fix version for this one at 2.2?
>> Yes, I am. I will try to find a fix as soon as possible. If it turns out
>> to be before the release of 2.1, it will go there. Otherwise, it will be
>> in 2.2.
>
> Thanks!
>
> Any other opinions on the issues that I left marked 2.1? If others
> are OK with this, I will move 321 and 351-352 to 2.2 and proceed to
> cut a 2.1 RC.
This seems fine to me.
Luc
>
> Phil
>> Dim.
>> ----------------------------------------------------------------------------
>>
>> Dimitri Pourbaix *
>> Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
>> CP 226, office 2.N4.211, building NO * and CARPE DIEM.
>> Universite Libre de Bruxelles *
>> Boulevard du Triomphe * Tel : +32-2-650.35.71
>> B-1050 Bruxelles * Fax : +32-2-650.42.26
>> http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
>> For additional commands, e-mail: dev-help@commons.apache.org
>>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Phil Steitz <ph...@gmail.com>.
Dimitri Pourbaix wrote:
> Phil,
>
>>> Right now, I see no option but avoiding the computation of A^tA. It is
>>> fair to ask for 2.2.
>>
>> OK, so you are OK with keeping the fix version for this one at 2.2?
>
> Yes, I am. I will try to find a fix as soon as possible. If it turns out
> to be before the release of 2.1, it will go there. Otherwise, it will be
> in 2.2.
Thanks!
Any other opinions on the issues that I left marked 2.1? If others
are OK with this, I will move 321 and 351-352 to 2.2 and proceed to
cut a 2.1 RC.
Phil
>
> Dim.
> ----------------------------------------------------------------------------
>
> Dimitri Pourbaix *
> Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
> CP 226, office 2.N4.211, building NO * and CARPE DIEM.
> Universite Libre de Bruxelles *
> Boulevard du Triomphe * Tel : +32-2-650.35.71
> B-1050 Bruxelles * Fax : +32-2-650.42.26
> http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Dimitri Pourbaix <po...@astro.ulb.ac.be>.
Phil,
>> Right now, I see no option but avoiding the computation of A^tA. It is
>> fair to ask for 2.2.
>
> OK, so you are OK with keeping the fix version for this one at 2.2?
Yes, I am. I will try to find a fix as soon as possible. If it turns out
to be before the release of 2.1, it will go there. Otherwise, it will be
in 2.2.
Dim.
----------------------------------------------------------------------------
Dimitri Pourbaix *
Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
CP 226, office 2.N4.211, building NO * and CARPE DIEM.
Universite Libre de Bruxelles *
Boulevard du Triomphe * Tel : +32-2-650.35.71
B-1050 Bruxelles * Fax : +32-2-650.42.26
http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Phil Steitz <ph...@gmail.com>.
Dimitri Pourbaix wrote:
> Hi,
>
>>> 28: there is still one spurious non zero singular value
>>> (3.547702387229884E-7 which I am going to try to kick out) instead of
>>> two.
>>>
>>> Dim.
>>
>> I pushed it to 2.2 because what remained to resolve did not look to
>> me like a showstopper for 2.1. I may be missing something, though,
>> regarding severity. I am fine getting this into 2.1 if you have a fix.
>
> I just checked the code and notice the problem is already present at the
> exit of TriDiagonalTransform. In order to compute SVD, one computes the
> eigen decomposition of A^tA which relies upon the Householder tri-diagonal
> transformation. The tri-diagonal maxtrix still has 3 non-zero main
> diagonal
> elements. The smallest one is about 1.e-13, i.e. slightly too large to be
> considered null against the largest, about 900.0
>
> Right now, I see no option but avoiding the computation of A^tA. It is
> fair to ask for 2.2.
OK, so you are OK with keeping the fix version for this one at 2.2?
Phil
>
> Regards,
> Dim.
> ----------------------------------------------------------------------------
>
> Dimitri Pourbaix *
> Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
> CP 226, office 2.N4.211, building NO * and CARPE DIEM.
> Universite Libre de Bruxelles *
> Boulevard du Triomphe * Tel : +32-2-650.35.71
> B-1050 Bruxelles * Fax : +32-2-650.42.26
> http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Ted Dunning <te...@gmail.com>.
See
http://www.amazon.com/Computations-Hopkins-Studies-Mathematical-Sciences/dp/0801854148for
the best treatment of this that I know.
This is particularly handy in converting Lanczos' algorithm from computing
partial eigenvector decompositions into an algorithm for partial SVD. This
adjoint approach is particular handy for sparse conditions because adjoining
with a zero matrix is pretty much free.
On Fri, Mar 19, 2010 at 4:12 AM, Dimitri Pourbaix
<po...@astro.ulb.ac.be>wrote:
> Thanks. Never seen that approach but it is mathematically equivalent
> indeed. I will try it.
>
Re: [math] Status of MATH-327?
Posted by Dimitri Pourbaix <po...@astro.ulb.ac.be>.
Ted,
> To get an SVD isn't it more common to compute the eigen decomposition of
>
> [ 0 A' ]
> [ A 0 ]
>
> Rather than A' A?
Thanks. Never seen that approach but it is mathematically equivalent
indeed. I will try it.
Regards,
Dim.
----------------------------------------------------------------------------
Dimitri Pourbaix *
Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
CP 226, office 2.N4.211, building NO * and CARPE DIEM.
Universite Libre de Bruxelles *
Boulevard du Triomphe * Tel : +32-2-650.35.71
B-1050 Bruxelles * Fax : +32-2-650.42.26
http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Ted Dunning <te...@gmail.com>.
To get an SVD isn't it more common to compute the eigen decomposition of
[ 0 A' ]
[ A 0 ]
Rather than A' A?
I have heard that this is supposed to avoid some problems with round-off
such as you are seeing. Moreover, many algorithms can be restated so that
this matrix never needs to be constructed explicitly.
On Wed, Mar 17, 2010 at 4:23 AM, Dimitri Pourbaix
<po...@astro.ulb.ac.be>wrote:
> I just checked the code and notice the problem is already present at the
> exit of TriDiagonalTransform. In order to compute SVD, one computes the
> eigen decomposition of A^tA which relies upon the Householder tri-diagonal
> transformation. The tri-diagonal maxtrix still has 3 non-zero main
> diagonal
> elements. The smallest one is about 1.e-13, i.e. slightly too large to be
> considered null against the largest, about 900.0
>
> Right now, I see no option but avoiding the computation of A^tA. It is
> fair to ask for 2.2.
>
Re: [math] Status of MATH-327?
Posted by Dimitri Pourbaix <po...@astro.ulb.ac.be>.
Hi,
>> 28: there is still one spurious non zero singular value
>> (3.547702387229884E-7 which I am going to try to kick out) instead of
>> two.
>>
>> Dim.
>
> I pushed it to 2.2 because what remained to resolve did not look to
> me like a showstopper for 2.1. I may be missing something, though,
> regarding severity. I am fine getting this into 2.1 if you have a fix.
I just checked the code and notice the problem is already present at the
exit of TriDiagonalTransform. In order to compute SVD, one computes the
eigen decomposition of A^tA which relies upon the Householder tri-diagonal
transformation. The tri-diagonal maxtrix still has 3 non-zero main diagonal
elements. The smallest one is about 1.e-13, i.e. slightly too large to be
considered null against the largest, about 900.0
Right now, I see no option but avoiding the computation of A^tA. It is
fair to ask for 2.2.
Regards,
Dim.
----------------------------------------------------------------------------
Dimitri Pourbaix *
Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
CP 226, office 2.N4.211, building NO * and CARPE DIEM.
Universite Libre de Bruxelles *
Boulevard du Triomphe * Tel : +32-2-650.35.71
B-1050 Bruxelles * Fax : +32-2-650.42.26
http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [math] Status of MATH-327?
Posted by Phil Steitz <ph...@gmail.com>.
Dimitri Pourbaix wrote:
> Hi,
>
> What is the present status of MATH-327? Phil updated on Saturday, I guess,
> to say it will be fixed in 2.2. However, the present version adresses the
> problem of the number of iterations reaching the maximum. It also improves
> (but does not completely solve) the situation mentioned by Luc in January
> 28: there is still one spurious non zero singular value
> (3.547702387229884E-7 which I am going to try to kick out) instead of two.
>
> Dim.
I pushed it to 2.2 because what remained to resolve did not look to
me like a showstopper for 2.1. I may be missing something, though,
regarding severity. I am fine getting this into 2.1 if you have a fix.
Phil
> ----------------------------------------------------------------------------
>
> Dimitri Pourbaix *
> Institut d'Astronomie et d'Astrophysique * Don't worry, be happy
> CP 226, office 2.N4.211, building NO * and CARPE DIEM.
> Universite Libre de Bruxelles *
> Boulevard du Triomphe * Tel : +32-2-650.35.71
> B-1050 Bruxelles * Fax : +32-2-650.42.26
> http://sb9.astro.ulb.ac.be/~pourbaix * mailto:pourbaix@astro.ulb.ac.be
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org