[Corpora-List] implement of Support Vector Machine(SVM) with weighting on each training data point

[Toby Hawker]

I believe there is an extension for libSVM that allows per-instance weights.

Link:
http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/#15

The zip file linked from there appears to be based on an older version
of libSVM, but you may be able to patch the latest code with similar
changes, or just use it as-is.

On Thu, Jul 31, 2008 at 1:22 PM, Xiaolin Wang
<arthur_general at sjtu.edu.cn> wrote:
> Hi,
>
>
>
>
>
> I'm looking for an available implement of Support Vector Machine, in which
> each training data point can be attached a weight value.
>
>
>
> I'm working on Adaboost recently, the base classifier of which must be
> trained on the dataset with a distribution of occurrence. Such distribution
> has exactly the same effect as weights which indicate how importance each
> training data point is.
>
>
>
> I used to use libSVM and SVMlight, but neither of them has such features.
> LibSVM only supports weighting on the whole class rather than data point.
> Anyone knows of such an implement?
>
>
>
>
>
> Thanks
>
>
>
> X.L. Wang

_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora