From: Lorenz B. <spo...@st...> - 2017-05-24 07:09:31
|
Without seeing the data + the used conf files for the it's almost impossible to see what's going on. At least for the ClassLearningProblem it should not return the target class itself. For others, indeed it will return the best solution regarding coverage - this might be also a simple class A if this is good enough to distinguish pos. from neg. examples. But as I said, we need data + learning setup to given you a better explanation. Cheers, Lorenz > > Hi, > > Thanks a lot for getting back to me. > > > When I'm using the Cross validation What I get as a result is the > Class A b= > y itself. > > > I did test it with both PosOnly learning problem and also PosNegative > learn= > ing problem. > > > considering the three different learning problem(lp1, lp2, lp3); > > > > * ClassLearningProblem lp1 =3D new ClassLearningProblem(rc); > * PosOnlyLP lp2 =3D new PosOnlyLP(rc, posExamples); > * PosNegLPStandard lp3 =3D new PosNegLPStandard(rc, > posExamples, negExamples); > > > Is that correct to conclude that the solution you get based on the lp1 > is more precise in comparison with lp2 and lp3? (maybe because the > number of th= > e negative instances are always more?) > > > I did work with the provided CrossValidation in DL code and there Only > PosNeg and PosOnly learning problems are supported, however all I get > as a solutionis either; A and (not B ) or A for each Target class. I > can not get anything to cover object relations. While if I ignore > cross-validation and only use lp1, I can have results including object > relations as well. > > > Best, > Faranak > > -- > > > > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > > _______________________________________________ > dl-learner-discussion mailing list > dl-...@li... > https://lists.sourceforge.net/lists/listinfo/dl-learner-discussion |