Learning Rules from Distributed Data