Open chupvl opened 6 years ago
It is correct that CrossConformalClassifier currently does not properly support conditional (Mondrian) Conformal Prediction.
You should, however, be able to construct an AggregatedConformalPredictor that uses Mondrian ICPs as the ensemble members, to achieve something similar:
model_adapter = ClassifierAdapter(SVC(probability=True))
nc = ClassifierNc(model_adapter)
icp = IcpClassifier(nc, condition=lambda x: x[1])
ccp = AggregatedCp(icp, CrossSampler())
Hello,
do I understand correctly that Mondrian Cross-Conformal Prediction is not implemented in the pipelines ?
Thanks!