[OTDev] Feature values
Nina Jeliazkova nina at acad.bgThu Dec 17 15:13:44 CET 2009
- Previous message: [OTDev] Feature values
- Next message: [OTDev] Feature values
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Dear Tobias, I've forwarded our recent discussion with Christoph on tuples + proposal in a previous mail. Tobias Girschick wrote: > Dear Christoph, > > I am not sure that I understand your question right. In your example, > are you talking about one feature that has a :smarts, p_value:, ....? > In this case we might have a (maybe) similar problem here, too. Because > at the moment I don't see a way to encode (descriptor calculation) > algorithm parameters in features. But if we generate descriptors by > pattern mining algorithms the algorithm parameters have to be encoded in > the feature somehow and the should be query-able to prevent double > calculations. > > IMHO this is a different case than Christoph's. A descriptor with different parameters is just a different Feature and should not be stored under the same feature URI. Then feature URI should have a way to denote it is calculated from the same algorithm (ot:hasSource was introduced to link to the Algorithm , or Model, or Dataset). Now we indeed have no way to include parameters in this setup. One idea is to consider descriptor calculation algorithms as Models (algorithm + specific parameters), and include link to the specific Model. (I think I've already suggested this some weeks before). Best regards, Nina > Regards, > Tobias > > > On Tue, 2009-12-15 at 14:24 +0100, Christoph Helma wrote: > >> Dear Nina, All, >> >> I have to represent feature values as tuples (i.e. name/value pairs, >> e.g. :smarts => "NN", :p_value => 0.97, :effect => "activating", or >> :prediction => "active", :confidence => 0.5). For this purpose I have >> invented the OWL Class "Tuple", and the ObjectProperty 'tuple' that can >> be used instead of a (feature)value. Do you think that tuples should be >> included into opentox.owl or are there any other solutions for >> representing complex features. >> >> Best regards, >> Christoph >> _______________________________________________ >> Development mailing list >> Development at opentox.org >> http://www.opentox.org/mailman/listinfo/development >> > > >
- Previous message: [OTDev] Feature values
- Next message: [OTDev] Feature values
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Development mailing list