11-745: Advanced Statistical Learning Seminar

Department: Language Technologies Institute (LTI)
Units: 12
Semester: Fall
Instructor: Yiming Yang

Prerequisites: Course 11-741 (Information Retrieval), 15-681 or 15-781 (Machine Learning) or consent of the instructor.

Textbook: "The Elements of Statistical Learning: Data Mining, Inference, and Prediction" by Trevor Hastie, Robert Tibshirani and Jerome Friedman

Course Description:
Advanced Statistical Learning Seminar is a course that emphasizes on the theoretical foundation of statistical learning and its applications to many challenging problems. The objective is to enhance the understanding of statistical methods that graduate students learned from different courses, including Machine Learning (15-681 or 15-781) and Information Retrieval (11-741), and to integrate scattered pieces of knowledge into a more comprehensive formulation.

The main theme in this course may vary to year. For Fall 2002, we choose the topics in the book "The Elements of Statistical Learning: Data Mining, Inference, and Prediction" by Trevor Hastie, Robert Tibshirani and Jerome Friedman. Specifically, the topics include both supervised learning and unsupervised learning issues, various linear regression methods, linear classification methods, basic expansions and regularization, kernel methods, model assessment and selection, model inference and averaging, boosting and additive trees, neural networks, support vector machines, nearest-neighbor methods, and unsupervised clustering. Additional topics may include computational geometry applied to machine learning problems and other issues.

The course will take the form of a seminar instead of lectures by the instructor. We will go through the book, one chapter per class except that a heavy chapter may be split into two classes. Each class starts by collecting questions from all the participants about the current chapter, followed by a presentation (lecture) on that chapter, and then classroom discussions about collected and new questions. Students will be grouped into teams of two or three persons; each team is assigned to two chapters (or three, depending on the enrollments) that they will analyze, deliver a lecture and lead the classroom discussions. All the students are required to read every chapter before it is discussed in a class, and present their questions at the start of the class.

Grading: There will be no exams and homework in this course. The grading will be based on participation in class, the quality of the seminar presentations (lectures) delivered by each team, and questions submitted at the start of each class.