mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Owen <sro...@gmail.com>
Subject Re: Embedding mahout in a java app
Date Wed, 02 Nov 2011 11:46:44 GMT
MahoutDriver is the closest thing to a single point of entry for all the
algorithms. It's for command line use but you can see what it does after
parsing args.

In general, most algorithms use Hadoop, so in general no there is not a
Hadoop free mode. Some bits have non Hadoop parts though that's the
exception.

Hadoop local mode works pretty fine, though it takes some work to package
it up for standalone use.
 On Nov 2, 2011 11:35 AM, "Tharindu Mathew" <mccloud35@gmail.com> wrote:

> Thanks Sean.
>
> Looks like I'll have to dig into the code will start from MahoutDriver.
>
> Is there a mode that will work for all algorithms. For example, all
> algorithms can run on a single node mode or all algorithms run on a hadoop
> mode ( I know Hadoop has a local mode, but that's not what I'm referring
> to) or something similar?
>
> I'd like to support the tool to run even without Hadoop as that will be
> great for small data sets for someone to try out and play around with.
> Maybe there's another java library that already does this.
>
> On Wed, Nov 2, 2011 at 4:51 PM, Sean Owen <srowen@gmail.com> wrote:
>
> > I see, the Java interfaces vary from area to area since different
> > algos are different things and sometimes take different input.
> >
> > Generally, the classifiers take in Mahout Vector input, and are
> > Hadoop-based, so you'd be writing some code to run Mahout jobs on
> > Hadoop from your GUI app. Not all are like this though.
> >
> > I don't think there's a one-stop easy interface already ready for you
> > here, no. You'd have to stitch together different parts of the code
> > and do some input transformation and Hadoop integration, I imagine.
> >
> > On Wed, Nov 2, 2011 at 11:17 AM, Tharindu Mathew <mccloud35@gmail.com>
> > wrote:
> > > I want to create a java UI tool (based on a web app) that can pick and
> > > apply different algorithms available in Mahout to different data sets.
> > >
> > > Hence the embedding with java. Obviously, I understand that everything
> is
> > > callable from Java since it's written in Java :).
> > >
> > > For example, I want to do a apply a classification (ex: Bayesian)
> > algorithm,
> > > and train on a data set stored in Cassandra. I don't expect a sample
> for
> > > Cassandra but at least a code sample that operates on a data set stored
> > csv
> > > file that applies an algorithm like Bayesian.
> > >
> > > I'd appreciate if you can point me to any code sample for this or
> > something
> > > similar?
> > >
> > > On Wed, Nov 2, 2011 at 3:32 PM, JAGANADH G <jaganadhg@gmail.com>
> wrote:
> > >
> > >> On Wed, Nov 2, 2011 at 2:51 PM, Tharindu Mathew <mccloud35@gmail.com>
> > >> wrote:
> > >>
> > >> > Hi Sean,
> > >> >
> > >> > I guess with a proper API it just makes it easier. I was hoping
> you'd
> > >> point
> > >> > me to a code sample or a tutorial.
> > >> >
> > >>
> > >>
> > >>
> > >> Hi
> > >>
> > >> For detailed code samples and tutorials see the book "Mahout in
> Action".
> > >> You will get a clear insight on how to use Mahout (in java in your
> case
> > >> !!!!)
> > >> --
> > >> **********************************
> > >> JAGANADH G
> > >> http://jaganadhg.freeflux.net/blog
> > >> *ILUGCBE*
> > >> http://ilugcbe.psgkriya.org
> > >>
> > >
> > >
> > >
> > > --
> > > Regards,
> > >
> > > Tharindu
> > >
> > > blog: http://mackiemathew.com/
> > >
> >
>
>
>
> --
> Regards,
>
> Tharindu
>
> blog: http://mackiemathew.com/
>

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