Implement a decision tree learner that handles input features with ordered domains. You can assume that any numerical feature is ordered. The condition should be a cut on a single variable, such as X ≤ v, which partitions the training examples according to the value v. A cut-value can be chosen for a feature X by sorting the examples on the value of X, and sweeping through the examples in order. While sweeping through the examples, the evaluation of each partition should be computed from the evaluation of the previous partition. Does this work better than, for example, selecting the cuts arbitrarily?