mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Dunning <ted.dunn...@gmail.com>
Subject Re: how to work with ARFF files using Mahout clustering
Date Sat, 15 Sep 2012 13:48:23 GMT
Mahout does not directly support SVM.

The classification framework allows additional loss functions to be
defined.  Using hinge loss with the current SGD framework should produce
results that are essentially equivalent to SVM.

You will still have the same issues that the current SGD framework imposes.
 These include:

- it is designed to work with sparse data

- it is single node and not easy to extend efficiently to map-reduce

- it doesn't implement solution averaging or conjugate gradient convergence
acceleration.


On Sat, Sep 15, 2012 at 6:13 AM, Rajesh Nikam <rajeshnikam@gmail.com> wrote:

> Hello,
>
> Mahout seems to have support for SVM as "Linear SVM Package (LIBLINEAR) for
> Mahout". However I dont see documentation and examples showning how to use
> SVM?
> What all kernels are supported ?
>
> Has someone done some experiments with its stability, memory usage and
> performance etc ?
>
>
> Looking forward for your reply.
>
> Thanks
> Rajesh
>
>
>
> On Fri, Sep 14, 2012 at 9:40 PM, Rajesh Nikam <rajeshnikam@gmail.com>
> wrote:
>
> > I see "Linear SVM Package (LIBLINEAR) for Mahout" however no
> documentation
> > found around this ?
> > Has someone done some experiments with it.
> >
> > Looking forward for your reply.
> >
> > Thanks
> > Rajesh
> >
> >
> >
> >  On Fri, Sep 14, 2012 at 9:31 PM, Rajesh Nikam <rajeshnikam@gmail.com
> >wrote:
> >
> >> Hi Ted,
> >>
> >> Sorry to bother you again.
> >>
> >> One quick question: Does Mahout support SVM, what is the Java class
> name ?
> >> Any inputs on its stability and performance ?
> >>
> >>
> >> Thanks
> >> Rajesh
> >>
> >>
> >>
> >>
> >> On Thu, Sep 13, 2012 at 8:53 PM, Ted Dunning <tdunning@maprtech.com
> >wrote:
> >>
> >>> Send this to the mailing list.
> >>>
> >>>
> >>> On Thu, Sep 13, 2012 at 7:35 AM, Rajesh Nikam <rajeshnikam@gmail.com
> >wrote:
> >>>
> >>>> Hi Ted,
> >>>>
> >>>> I have data in WEKA ARFF format.
> >>>>
> >>>> What to how to use this ARFF formatted data is to be clustered using
> >>>> mahout clustering ?
> >>>>
> >>>> I have used following command to covert ARFF into vector file.
> >>>>
> >>>> bin/mahout arff.vector -d test.arff -t dict.txt -o kmeans/seqfiles
> >>>>
> >>>> How to use generated mvc file for clustering algorithms like canopy
or
> >>>> kmeans ?
> >>>>
> >>>> How to split generated clusters ?
> >>>>
> >>>> Thanks,
> >>>> Rajesh
> >>>>
> >>>>
> >>>
> >>
> >
>

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