Base learner implementation and parameters
For the simulation experiment, we tested three base learners: ExtraTreeRegressor, Linear Regression and Support Vector Machines (SVM). For ExtraTreesRegressor, we used 1000 trees with a maximal depth of 3, squared error as the criterion of split, minimal number of examples as 2 in a split and minimal number of examples in a node as 1. For SVM, we used the SVR (support vector regressor) implemented in sklearn, with C=1.0, and epsilon = 0.2. For Linear Regression we used Ridge penalization with alpha = 1.0.