Pantelis, All,

chung wrote:
> Dear all,
>   According to the OpenTox specifications for Algorithm
> ( http://opentox.org/dev/apis/api-1.0/Algorithm  ) if a client needs to
> train a classification or a regression model, then a dataset uri and
> some tuning parameters for the algorithm have to be provided. As I
> mentioned before, the target feature has also to be specified by the
> client. 
>
>   
Target feature as well as predictors are expected to be included in the
dataset. What is currently indeed missing is a meta information,
specifying which features are predictors and which are target (can be
more than one).
> Just for clarity, I'd like to ask whether we should consider that the
> descriptor calculation is already done or not. For example, if we find
> the compound "/compound/101" and the feature definition
> "/feature_definition/1234" in a given data set, then is there a value at
> "/feature/compound/101/feature_definition/1234" or we have to call the
> descriptor calculation web service to calculate the feature and generate
> this URI? Should we expect a status code 404 from
>   

> "/feature/compound/101/feature_definition/1234"?
>
>   
Both could happen. Perhaps a reasonable workflow is to verify if the
value(s) exist and then initiate their generation/retrieval.  IMHO 404
should be expected to be returned by any of the resources.

Best regards,
Nina
> Best regards,
> Pantelis.
>  
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>