This chapter proves how machine learning can succeed. By making some assumptions, we will see that machine learning can learn a prediction rule (function) that makes small errors on real world data as well as on training data.