Sure, I could do that. Would need a couple days before I get to it,
though...
Regarding your statement that interpolated precision makes only sense on
ranked results, are you saying that whenever one is trying to build a
non-IR classifier then one should always go with ROC curves? I find myself
using PR curves more often then ROC curves and then there is always the
question why one would choose a certain operation point when there is
another one that promises better precision at higher recall.
2013/3/16 Lars Buitinck <[email protected]>
> 2013/3/16 Lars Buitinck <[email protected]>:
> > Interpolated precision only makes sense on ranked results. We don't
> > have any ranking models.
>
> ... on second thought, precision_recall_curve could be interpreted as
> doing some kind of ranking evaluation (if the probabilities are used
> to rank outputs), so maybe an interpolate_precision argument might be
> in order. Care to submit a patch and a nice graph to demo it in the
> docs?
>
> --
> Lars Buitinck
> Scientific programmer, ILPS
> University of Amsterdam
>
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_d2d_mar
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_mar
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
