ECE News Story

Winter 2014: EECS 598-005 Statistical Learning Theory

Course No.: EECS 598-005
Credit Hours: 3 credits
Instructor: Clayton Scott
Prerequisites: EECS 501 or equivalent

Course Description:
In this course we will prove performance guarantees that quantify the ability of a machine learning algorithm to generalize from training data to unseen test data. Potential topics to be covered include concentration of measure, uniform deviation bounds, empirical and structural risk minimization, Rademacher complexity, Vapnik-Chervonenkis theory, consistency and rates of convergence, margin-based bounds, stability bounds, and application of these theories to learning algorithms such as decision trees, boosting, support vector machines, and kernel density estimators.
[More Info]