commons-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Dunning <ted.dunn...@gmail.com>
Subject Re: [Math] compare/complement r
Date Fri, 30 Apr 2010 00:50:20 GMT
Do you mean you are writing C++ to be called from Java?

If so, one of Commons Math, Mahout <http://lucene.apache.org/mahout/> (with
matrix operations derived distantly from Colt) or
JMT<http://code.google.com/p/matrix-toolkits-java/> are
likely to fit your needs much better.

None of these systems satisfies all needs, but where they shine you would be
hard-pressed to beat them in C++ with your own code.  In some cases, there
exist faster C++ codes, but if you are writing your own you aren't there.

On Thu, Apr 29, 2010 at 5:43 PM, Michael Stover
<michael.r.stover@gmail.com>wrote:

> As for scaling, we have needs to scale from a single desktop to using MPI
> in
> a supercomputer center.  We are currently writing all our own math code in
> C++ that does a lot of what R and commons/math does, but with a completely
> different API (of course).  The code is crap, and I am arguing to use
> existing java libraries, but I need to be able to answer their questions
> about performance and scalability.
>
> -Mike
>
> On Thu, Apr 29, 2010 at 9:26 AM, adasal <adam.saltiel@gmail.com> wrote:
>
> > It probably v. good, but exactly? Also how being used, I don't know how
> you
> > would scale R, but you may wish to scale your usage of commons. There is
> > the
> > issue of notation, I don't think Java is very maths friendly, at least I
> > don't find it so.
> >
> > Adam
> >
> > On 29 April 2010 14:10, Michael Stover <michael.r.stover@gmail.com>
> wrote:
> >
> > > And does the performance compare between R and commons math?
> > >
> > > On Thu, Apr 29, 2010 at 7:09 AM, Sachin Dole <sachin.dole@gmail.com>
> > > wrote:
> > >
> > > > Thank you! I reached a similar inference reading up on the site and
> now
> > > > after reading this email I feel like I have a strong confirmation.
> > Thanks
> > > > again.
> > > >
> > > > On Apr 28, 2010 9:11 PM, "Phil Steitz" <phil.steitz@gmail.com>
> wrote:
> > > >
> > > > Sachin Dole wrote:
> > > > > R probably has a large superset of features that math provides
> while
> > it
> > > > > tries...
> > > > You are correct that in general R provides a superset of what
> > > > commons math does, though there are a few things that commons math
> > > > provides that R does not.  There is a lot of overlap and in many
> > > > cases the functionality that is provided by commons math is similar
> > > > to what R provides, though of course the APIs are different. We test
> > > > some of the commons math implementation classes against R (see the R
> > > > subdirectory in src/test).  Commons math will not be of much value
> > > > as a wrapper / invocation framework for R; but it can be used
> > > > directly to do some of the same computations that R does.  This was
> > > > part of the original motivation for creating commons math.
> > > >
> > > > The best way to get an overview of what is provided by commons math
> > > > is to look at the User Guide:
> > > > http://commons.apache.org/math/userguide/index.html
> > > >
> > > > Phil
> > > >
> > > > >
> > > > > On Apr 28, 2010 6:22 AM, "Rory Winston" <rory.winston@gmail.com>
> > > wrote:
> > > > >
> > > > > Sachin
> > > > >
> > > > > Common...
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message