You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by Jeff Eastman <je...@windwardsolutions.com> on 2008/03/18 20:00:07 UTC

RE: ´ð¸´: input and output format following PMML specification

+1 on the benefits of interoperability

How would you go about identifying and selecting from the alternatives? What decision criteria would you suggest? What other alternatives should be considered? Go ahead and open a Jira if you want to lead this discussion.

Jeff

> -----Original Message-----
> From: shunkai.fu [mailto:shunkai.fu@roboo.com]
> Sent: Monday, March 17, 2008 5:42 PM
> To: mahout-dev@lucene.apache.org
> Subject: 答复: input and output format following PMML specification
> 
> Hi,
> 
> I am new to this group, so I may miss some. However, I still feel happy to
> discuss with you guys to make things better.
> 
> A standard presentation has the following benefit:
> 1) What is learned via Mahout can be recognized somewhere using other
> software, for viewing, updating and predicting purpose;
> 2) What is learned using other ML or DM software can be applied the
> similar
> processing on Mahout;
> 
> By the way, some models support updating with additional data, so we may
> need to put this into consideration. Different learning from ground, the
> input here is likely be a model described by XML.
> 
> Thanks,
> 
> Shunkai
> 
> -----邮件原件-----
> 发件人: Grant Ingersoll [mailto:gsingers@apache.org]
> 发送时间: 2008年3月18日 0:28
> 收件人: mahout-dev@lucene.apache.org
> 主题: Re: input and output format following PMML specification
> 
> Is this also overlapping w/ what Karl suggested earlier using the Data
> Mining JSR?
> 
> On Mar 17, 2008, at 11:55 AM, Jeff Eastman wrote:
> 
> > I think the whole area of intra-Mahout and extra-Mahout
> > representations is
> > one that needs further study. I glanced over the URL you provided
> > and this
> > certainly looks like something to consider. It is a bit premature to
> > pick
> > one at this point, but please feel free to investigate how it might be
> > applied to Mahout. In terms of practical next steps you could:
> >
> > 1. Open a Jira to investigate external representation formats for
> > Mahout
> > 2. Propose a set of criteria that you think ought to apply to such
> > representations
> > 3. Contribute a business case for why this particular representation
> > would
> > satisfy those criteria
> > 4. Investigate alternative representations and evaluate them so we
> > can be
> > sure we have done a reasonable due diligence
> >
> > These steps would put us all in a much better position to answer your
> > original question.
> >
> > Jeff
> >
> >> -----Original Message-----
> >> From: shunkai.fu [mailto:shunkai.fu@roboo.com]
> >> Sent: Sunday, March 16, 2008 6:24 PM
> >> To: mahout-dev@lucene.apache.org
> >> Subject: input and output format following PMML specification
> >>
> >> Hi,
> >>
> >>
> >>
> >> Do you guys think we need support PMML format, in term of input and
> >> output?
> >>
> >>
> >>
> >> It is proposed by several key companies focusing on DM, and you can
> >> find
> >> more information on http://www.dmg.org/ .
> >>
> >>
> >>
> >> Best,
> >>
> >>
> >>
> >> Shunkai
> >>
> >>
> >
> >
> >
> 
> --------------------------
> Grant Ingersoll
> http://www.lucenebootcamp.com
> Next Training: April 7, 2008 at ApacheCon Europe in Amsterdam
> 
> Lucene Helpful Hints:
> http://wiki.apache.org/lucene-java/BasicsOfPerformance
> http://wiki.apache.org/lucene-java/LuceneFAQ
> 
> 
> 
>