mahout-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Dunning <ted.dunn...@gmail.com>
Subject Re: SVM implementation for Mahout
Date Mon, 04 May 2009 04:39:45 GMT
I think that the primary criterion is that we want to provide capabilities
for users of Mahout.  Providing an SVM implementation would be the first
priority, regardless of source (presuming licensing and such is handled
correctly).

That said, the rub is likely to come when a scalable implementation is to be
produced.  SVM isn't nearly as simple to parallelize as some other
algorithms so it may well become necessary to make some pretty substantial
modifications.

In the meantime, whatever is the easiest way to make forward progress is
probably your best bet.

On Sun, May 3, 2009 at 7:45 PM, Joe Kumar <joekumar@gmail.com> wrote:

> I just got back from Vikas reg svmlin and he is ok to use apache 2.0
> license.
>
> That aside, I thought of sharing some of my perspectives.
> SVM is one of the widely used classification algorithms and I guess it is
> better to have our own implementation. We could reuse existing libraries
> for
> non-core algorithms or components. Assuming that SVM on Mahout would be
> quite widely used, would it be better to write our own ?
>
> On Mon, Apr 6, 2009 at 1:58 AM, Sean Owen <srowen@gmail.com> wrote:
>
> > Apache 2.0 License would be easiest I imagine?
> >
> > On Mon, Apr 6, 2009 at 6:12 AM, Joe Kumar <joekumar@gmail.com> wrote:
> > > He was asking about what licensing he should specify for svmlin so that
> > it
> > > can used in mahout ? any thoughts / suggestions please
> >
>



-- 
Ted Dunning, CTO
DeepDyve

111 West Evelyn Ave. Ste. 202
Sunnyvale, CA 94086
www.deepdyve.com
858-414-0013 (m)
408-773-0220 (fax)

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