On Mon, Jan 2, 2012 at 11:45 AM, Nina Jeliazkova
<jeliazkova.nina at gmail.com>wrote:

> Hi Martin,All,
>
> Happy new year to all!
>
> On 2 January 2012 12:24, Martin Guetlein <martin.guetlein at googlemail.com
> >wrote:
>
> > Hi all,
> >
> > First of all, I wish you all a happy new year 2012.
> >
> > We want to use generic OpenTox prediction models. Alongside the normal
> > prediciton value, we want to get a measure of how well the predicted
> > compound fits the AD of training dataset (similar to the lazar
> confidence).
> >
> > Did someone try this before?
> > @Nina, is it possible to use this service for that:?
> > http://apps.ideaconsult.net:8080/ambit2/algorithm/superservice
>
>
> Yes, it should be possible. AD are regular OpenTox algorithms/models, just
> replace the model URI with AD model URI.
>


Great, I'll give it a try.


>
>
> >
> >
> > In case there is nothing availble yet, I would try to wrap a prediction
> > algorithm and one of ambits AD algorithms:
> >
> >
> http://apps.ideaconsult.net:8080/ambit2/algorithm?type=http%3A%2F%2Fwww.opentox.org%2FalgorithmTypes.owl%23AppDomain
> > Any thoughts / opinions on this approach?
> >
>
> Fine, if it suits your purposes.
>
> Just to remind AD algorithms here are not dependent of the model itself,
> only on the training set.  E.g. having trained an AD algorithm with a
> given dataset, you'll get the same results , regardless of the model used.
>

Sure. The AD algorithm and the prediction algorithm have to use the
identical features, otherwise it makes no sense.

Best regards,
Martin



>
> Best regards,
> Nina
>
>
> >
> > Best regards,
> > Martin
> >
> > --
> > Dipl-Inf. Martin Gütlein
> > Phone:
> > +49 (0)761 203 7633 (office)
> > +49 (0)177 623 9499 (mobile)
> > Email:
> > guetlein at informatik.uni-freiburg.de
> > _______________________________________________
> > Development mailing list
> > Development at opentox.org
> > http://www.opentox.org/mailman/listinfo/development
> >
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>



-- 
Dipl-Inf. Martin Gütlein
Phone:
+49 (0)761 203 7633 (office)
+49 (0)177 623 9499 (mobile)
Email:
guetlein at informatik.uni-freiburg.de