Statistical learning theory provides the general framework for analyzing the performance of supervised learning algorithms. In this talk, I will illustrate some basic concepts that are essential for the study of learning rate of an algorithm, using a simple example, where a weird distribution is successfully learned by a surprisingly simple algorithm. And at the end, I will briefly discuss the implication of Devroye's No Free Lunch Theorem, and how the fast learning rate is possible for kernel support vector machine. Speaker(s): Yitong Sun (University of Michigan)