View selection in multi-view stacking: Choosing the meta-learner