public interface MultiTargetClassifier
extends weka.classifiers.Classifier, weka.core.OptionHandler
MultiTargetClassifier.java - A Multi-Target (i.e., Multi-Output / Multi-Dimensional) Classifier.
Implementing this interface signals to the Evaluation that we are dealing with multi-target data,
and a different evaluation output is given. Training and classification is the same, using the
methods buildClassifier(Instances) and distributionForInstance(Instance), except that
the latter only returns the argmax value (i.e., what is to be considered the predicted value).
At the moment it is also possible to extend the double from distributionForInstance
to a vector of L*2 doubles instead of L which contain the max.
In other words, k in position j and p(y[j]=k) in position j+L.
In the future we will make use of double distributionForInstance(Instance,int) instead.