Hi Chris.
Thanks for your offer to contribute.
I'm not familiar with the paper. Is this a generic way to do the
multiclass -> binary transformation?
Then this might be a good addition to the OneVsRestClassifier.
I can not judge how relevant the technique is in practice, though.
Do you (or anyone else) have any experience?
Best,
Andy
On 12/06/2012 03:22 PM, Chris Brew wrote:
Would people be interested in having the cost-sensitive one-versus-all
reduction from Vowpal Wabbit in sklearn,multiclass?
cf. http://hunch.net/~beygel/woa.pdf <http://hunch.net/%7Ebeygel/woa.pdf>
It looks to me like a feasible thing to implement, given the
description in the paper, the implementation in vw and the existing
code in sklearn, To do this I would use the sample_weight parameter to
the fit method on the underlying binary classifier, which would limit
applicability to the classifiers that support that.
I would be a first-time contributor, so I might be underestimating
what is involved. What do people think?
Chris
--
Chris Brew, Educational Testing Service
------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
