Learning Priorities From Noisy Examples