Support for parallelism in ensemble learning

Ensemble learning techniques take standard machine learning methods and improve their performance. This is achieved by learning a set of models and combining their predictions in some fashion. Ensemble learning methods whose constituent classifiers are independent from one another, in the sense that the input to each classifier doesn't depend on the output of any other classifier(s), lend themselves to parallel construction. Ensemble learning methods in Weka that can be made parallel in such a fashion include:

• weka.classifiers.meta.Bagging
• weka.classifiers.meta.RandomCommittee
• weka.classifiers.meta.RandomSubSpace
• weka.classifiers.meta.RotationForest
• weka.classifiers.meta.Vote
• weka.classifiers.meta.Stacking
• weka.classifiers.meta.MultiScheme

Support for parallel processing on multi cpu/core machines for these methods will be available in Weka 3.7.1.