Learning by Leaps and Bounds