EE511 -- Introduction to Statistical Learning

Course Information, Spring 2008

Announcements:

• Project evaluation data is now posted.
• Ostendorf office hours this week are move to Monday 6/2 1:30-3:30.
• Final presentations will be June 10, starting at 3:30 in rm EE403.

Course Hours: T,Th 2:30-4:20
Location: EEB 025

Lecturers:
Mari Ostendorf (mo@ee)
Office: EE Rm 215D (inside 215, door by 205)
Office Hours: Weds 3:30-5:00
Kevin Duh (kevinduh@u)
Office Hours: Tu 4:30-5:30 (EEB M306)

Grader: Anna Margolis (amargoli@u)

Course description:
The course covers classification and estimation of vector observations, including both parametric and nonparametric approaches. Topics include classification with likelihood functions and general discriminant functions; density estimation; supervised, semi-supervised and unsupervised learning; feature reduction and selection; model selection; and performance estimation. Important classes of models are covered, including Gaussian mixtures, decision trees, SVMs, and neural networks. The course will have one midterm and a final project involving classification of data delivered early in the term. (Offered alternate years.)

Course objectives:
Students who complete this course should gain:

• knowledge of the basic theory of different methods for statistical learning;
• an understanding of the trade-offs of different approaches and assumptions under which different approaches are equivalent;
• intuition for the problems of data sparsity and overtraining, and for choosing among options for dealing with this problem; and
• practical experience using different approaches and appropriate evaluation methodologies.

Pre-requisite: Introductory Probability and Statistics (EE505 or undergrad equivalent)

Grading Policy
Homework: 40%
Exam: 30% (May 20, 2008)
Final Project: 30% (see Final Project Guidelines), Final Presentation June 10, 3:30-6:30

Grader: TBD

• Schedule

• HWs and Handouts

• Collaboration and Make-up Policies

• Information on EE Computer Access