Practical PAC Learning

New strategies for "probably approximately correct" (par) learning that use fewer training examples than previous approaches are presented, demonstrating that pac-learning can be far more efficiently achieved in practice than previously thought.