Hi Nina, Martin, All,

On Wed, 2009-12-09 at 17:53 +0100, Martin Guetlein wrote: 
> Hi Nina, Tobias, All,
> 
> > I would like to post few questions for the clarification of the Algorithm -
> > Model - Validation relationship:
> >
> > ...
> >
> > - Does the Validation service rely on such functionality, or calculates the
> > relevant statistics itself ?
> >
> 
> No, here is my (validation) point of view:
> 
> Prediction models do not calculate prediction statistics on their own,
> they just make predictions.
> You can use the validation service to get the RMSE of a regression
> model on some dataset. The corresponding REST operation is the third
> command in the API
> (http://www.opentox.org/dev/apis/api-1.1/Validation). The RMSE as well
> as other regression statistics are available in the validation object
> which is returned.

Fabian and I have thought about that and we think that the fact that our
algorithms expose supported statistics might be due to the fact we
normally work. Using the same implementation for model building and
validation. But in OpenTox that is not the case. We have separate
services for model building and validation. Consequently, the algorithms
building the model does not have to support any statistics. They are as
Martin points out correctly calculated in the validation service. The
only thing that has to be exposed (maybe) is, what kind of statistics
can be calculated on the predictions. Classification or regression
stats.
So in our opinion we could get rid of the "statisticsSupported" stuff.
We may expose if the algorithm is classification or regression but that
should be covered by the algorithm ontology. 

Opinions?

best regards
Tobias

> 
> Best regards,
> Martin
> 
> 
> > Best regards,
> > Nina
> >
> 
> 
> 
> --
> Dipl-Inf. Martin Gütlein
> Phone:
> +49 (0)761 203 8442 (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


-- 
Dipl.-Bioinf. Tobias Girschick

Technische Universität München
Institut für Informatik
Lehrstuhl I12 - Bioinformatik
Bolzmannstr. 3
85748 Garching b. München, Germany

Room: MI 01.09.042
Phone: +49 (89) 289-18002
Email: tobias.girschick at in.tum.de
Web: http://wwwkramer.in.tum.de/girschick