Scalable algorithms for learning large decision trees