Hi Christoph,
 Thanks. I will send an email to Martin. Here are some examples with
curl:

1. Train an MLR model:

curl -X POST -d
'dataset_uri=http://ambit.uni-plovdiv.bg:8080/ambit2/dataset/6?max=100&prediction_feature=http://ambit.uni-plovdiv.bg:8080/ambit2/feature/11954' http://opentox.ntua.gr:3000/algorithm/mlr

2. Train a Naive Bayes Classification Model:

curl -X POST -d
'dataset_uri=http://ambit.uni-plovdiv.bg:8080/ambit2/dataset/9?max=100&prediction_feature=http://ambit.uni-plovdiv.bg:8080/ambit2/feature/12132' http://opentox.ntua.gr:3000/algorithm/naiveBayes

3. Train a Support Vector Machine Model for classification purposes:

curl -X POST -d
'dataset_uri=http://ambit.uni-plovdiv.bg:8080/ambit2/dataset/9?max=100&prediction_feature=http://ambit.uni-plovdiv.bg:8080/ambit2/feature/12132' http://opentox.ntua.gr:3000/algorithm/svc

4. Train a (Regression) Support Vector Machine Model:

curl -X POST -d
'dataset_uri=http://ambit.uni-plovdiv.bg:8080/ambit2/dataset/9?max=100&prediction_feature=http://ambit.uni-plovdiv.bg:8080/ambit2/feature/12132' http://opentox.ntua.gr:3000/algorithm/svc

5. Get the independent/dependent/predicted variables of a model:

curl http://opentox.ntua.gr:3000/model/12345/independent 
curl http://opentox.ntua.gr:3000/model/12345/dependent 
curl http://opentox.ntua.gr:3000/model/12345/predicted

(The results are available in both RDF and text/uri-list )

6. Use a model for prediction, providing a dataset which contains all
independent variables of the model:

curl -d 'dataset_uri=http://ambit.uni-plovdiv.bg:8080/ambit2/dataset/6'
http://opentox.ntua.gr:3000/model/58826

Within the next few days our models will be incorporated in ToxPredict
so predictions will be easy to perform.

Best Regards,
Pantelis


On Tue, 2010-03-02 at 10:55 +0100, Christoph Helma wrote:
> Dear Pantelis,
> 
> Can you get into contact with Martin and ask him to try his validation
> service with your algorithms? As soon as this works I can start to
> add your algorithms to ToxCreate. Can you also send me a few examples (or
> send me a documentation link) for model creation?
> 
> Thanks,
> Christoph
> 
> Excerpts from GitHub's message of Mon Mar 01 18:44:50 +0100 2010:
> > Hi Christoph, 
> >  We have some algorithms which seem to work without bugs right now. These are
> > mlr, svm, naiveBayes and svc. These algorithms can be integrated in ToxCreate
> > but some of them (in particular svm and svc) accept some optional parameters as
> > well which can lead to different models. So, I think that the user interface
> > will need some extensions.
> > 
> > View this Issue online: http://github.com/helma/opentox-toxmodel/issues#issue/3
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>