Random Forest Distillation

Let's approximate some random forests in a single decision tree. We greedily minimize the L2 distance between the decision tree and the random forest without access to data.