You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@commons.apache.org by Daniela Kolarova <da...@gmail.com> on 2009/02/24 07:37:41 UTC
[MATH] Current stage of the commons-math.linear package
Hello All,
I am interested in the develoment of the linear algebra subpackage of the
commons-math project. Can someone give me infomation on the current state of
the package and the TODO lists, plannings for the near future?
Thanks in advance!
Re: [MATH] Suggesting new package commons-math.simulation
Posted by Luc Maisonobe <Lu...@free.fr>.
In addition to the points identified by Siegfried, could you open a JIRA
issue for this ? You can do this by registering into
https://issues.apache.org/jira/secure/Dashboard.jspa and then select the
link "create a new issue in project Commons Math" that will appear in
the page https://issues.apache.org/jira/browse/MATH. Describe your
request and once the issue is created you will be able to attach files
to it. This is the proper way to provide patches.
Also if the code is more than a few lines long, which is almost
certainly the case for Monte-Carlo method, we will need a signed
Software Grant (see http://www.apache.org/licenses/#grants). For
discussions about these legal aspects, feel free to contact me directly
off list.
Did you write everything yourself or did you get inspiration from
existing published code ?
Thanks for contributing
Luc
Siegfried Goeschl a écrit :
> Hi Hicham,
>
> I'm not involved in commons-math but a few general tips ... some obvious
> and some not so obvious
>
> 1) is the code ASF licence compatible, e.g. not having GPL snippets or
> other license issues
> 2) each files must contain an Apache License Header
> 3) you have to follow the coding and naming conventions used in commons-math
> 4) proper documentation of all methods and fields
> 5) some JUnit tests to prove that the code is correct
>
>>>From that point on you can integrate your stuff locally in the
> commons-math build and create a patch file ...
>
> And you have to do 5) - really - if a committer takes the time to apply
> a non-trivial patch everything should go smoothly. If not the risk is
> too big to break the existing code base and you never know when this guy
> has another day to spare for your patch. And after a few weeks the code
> and your patch is out-of-sync causing even more work and delay .....
>
> Cheers,
>
> Siegfried Goeschl
>
>
> Hicham Wahbi wrote:
>> Hello,
>>
>> I am interested in contributing to a "commons-math. Simulation" package
>> since the Monte-Carlo estimation method is listed as an item in the wish
>> list.
>> This is the first time I try to contribute to an open source project. I
>> actually have already written the code and just need to submit it for
>> discussion. Please help me jump start.
>>
>> Thanks,
>>
>> Hicham Wahbi
>>
>>
>> ---------------------------------------------------------------------
>> 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] Suggesting new package commons-math.simulation
Posted by Siegfried Goeschl <si...@it20one.at>.
Hi Hicham,
I'm not involved in commons-math but a few general tips ... some obvious
and some not so obvious
1) is the code ASF licence compatible, e.g. not having GPL snippets or
other license issues
2) each files must contain an Apache License Header
3) you have to follow the coding and naming conventions used in commons-math
4) proper documentation of all methods and fields
5) some JUnit tests to prove that the code is correct
>From that point on you can integrate your stuff locally in the
commons-math build and create a patch file ...
And you have to do 5) - really - if a committer takes the time to apply
a non-trivial patch everything should go smoothly. If not the risk is
too big to break the existing code base and you never know when this guy
has another day to spare for your patch. And after a few weeks the code
and your patch is out-of-sync causing even more work and delay .....
Cheers,
Siegfried Goeschl
Hicham Wahbi wrote:
> Hello,
>
> I am interested in contributing to a "commons-math. Simulation" package
> since the Monte-Carlo estimation method is listed as an item in the wish
> list.
> This is the first time I try to contribute to an open source project. I
> actually have already written the code and just need to submit it for
> discussion. Please help me jump start.
>
> Thanks,
>
> Hicham Wahbi
>
>
> ---------------------------------------------------------------------
> 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
[MATH] Suggesting new package commons-math.simulation
Posted by Hicham Wahbi <hi...@hotmail.com>.
Hello,
I am interested in contributing to a "commons-math. Simulation" package
since the Monte-Carlo estimation method is listed as an item in the wish
list.
This is the first time I try to contribute to an open source project. I
actually have already written the code and just need to submit it for
discussion. Please help me jump start.
Thanks,
Hicham Wahbi
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [MATH] Current stage of the commons-math.linear package
Posted by Luc Maisonobe <Lu...@free.fr>.
Daniela Kolarova a écrit :
> Hello All,
>
> I am interested in the develoment of the linear algebra subpackage of the
> commons-math project. Can someone give me infomation on the current state of
> the package and the TODO lists, plannings for the near future?
That's great, thanks for your interest.
The package has been recently vastly improved for the upcoming 2.0
version. You can find the current in the subversion repository in trunk.
For now, the package has the following features:
- Real dense matrices with two different layouts (straightforward
double[][] array for simple cases and blocks layout for cache
efficiency with large matrices)
- Big decimal real matrices
- Real sparse matrices
- Real dense vectors
- Real sparse vectors
- decomposition algorithms
- Cholesky
- LU
- QR
- eigen decomposition
- singular value decomposition
- exact and least square solvers for AX=B using the previous
decomposition algorithms
Features that can be improved or added are:
- sparse matrices multiplication is not efficient
(see http://issues.apache.org/jira/browse/MATH-248)
- eigen decomposition is currently limited to symmetric matrices
(see http://issues.apache.org/jira/browse/MATH-235)
- the decomposition algorithms should be also provided in an
additional version supporting some new Field class for elements
(see http://markmail.org/thread/h2rmq6wo7nwqzx4w)
- the singular value decomposition uses only the lower level parts of
the very efficient MRRR algorithm (i.e. the dqd/dqds algorithms), it
should implement the complete algorithm
- an attempt to use recursive layout with variable size blocks for
matrices to improve again DenseRealMatrix has been made but failed,
the speed improvement was poor (less than 5% and only for quite large
matrices) and the class is not completed (see
http://markmail.org/message/nol6a3efb56zb5yk)
- LU decomposition is not optimized at all, it could probably be
improved by taking care of cache or using block algorithms (see
http://www.tau.ac.il/~stoledo/Pubs/lu.ps.gz)
Luc
>
> Thanks in advance!
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org