[prev in list] [next in list] [prev in thread] [next in thread] 

List:       wekalist
Subject:    Re: [Wekalist] multilabel classification
From:       "Iryna Skrypnyk" <iryna () cs ! jyu ! fi>
Date:       2004-05-22 21:30:59
Message-ID: 006301c44048$cdcacee0$fdb2ea82 () iryna
[Download RAW message or body]

Hello,

I have raised the similar question in the WEKA list some time ago.
For those who is interested the example of a multilabel classification
problem and SVM method for it is considered here
http://www1.cs.columbia.edu/compbio/papers/exp-phylo.html

Regards,
Iryna

----- Original Message ----- 
From: "Robert Strötgen" <stroetgen@rz.uni-hildesheim.de>
To: <wekalist@list.scms.waikato.ac.nz>
Sent: Thursday, May 20, 2004 10:57 AM
Subject: Re: [Wekalist] multilabel classification


> Hi Arturo,
>
> thanks for your answer!
>
> > First, study the imbalance of your classes. Usually you can gain in
speed
> > by just not considering those that are rare and seldom represented (to
me,
> > just training 200 out from 1000 is good enough without loosing to much
in
> > precision and recall).
>
> I already do this. My 1000 class labels are 1000 out from about 5000 ...
>
> Regards,
> Robert. :)
>
> -- 
> ~~~~~
> Robert Strötgen
> Informationswissenschaft
> IFAS, Universität Hildesheim * Marienburger Platz 22 * 31141 Hildesheim
> Tel.: +49 - 5121 - 883 847 * Fax:  +49 - 5121 - 883 802
> http://www.uni-hildesheim.de/~stroetge/
>
>
>
> _______________________________________________
> Wekalist mailing list
> Wekalist@list.scms.waikato.ac.nz
> https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist


_______________________________________________
Wekalist mailing list
Wekalist@list.scms.waikato.ac.nz
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist

[prev in list] [next in list] [prev in thread] [next in thread]