Learning Algorithms in the Limit