Learning Relational Rules from Rewards - World Wide