Abstract
In this article, an overview of basic tools in statistical learning theory is given. The aim of learning theory is to clarify the meaning of a learning method, justify the method and show an optimality of that. In particular, analyzing the behavior of generalization error is one of the most important issues. To analyze it, the empirical process theory plays a vital role. Technical tools such as Rademacher complexity, covering number and Dudley's integral are useful in the analysis. Finally, minimax optimality is discussed. A theoretic technique to give a lower bound of the minimax risk is presented.
