In this repository, we implemented several methods to MERGE decision trees.
The merge function provides the scalability to scale Decision Tree training into hundreds or more machines and then merge them back into a reasonable size.
With this approach, we developed the federated decision tree with about the same performance as boosted trees, while maintaining a size upper bound with feature number. This greatly increase the scalability of training decision trees on separated machines.