|
Course description: Machine learning is the study of computational methods that allow a machine to learn, that is, to improve its performance with experience. This is the second in a two-course sequence of graduate-level courses in machine learning. Topics to be covered include: classification and regression trees, ensemble methods, regression, elements of computational learning theory, support vector machines, hidden Markov models, expectation maximization, Markov chain Monte Carlo, constructive methods, relational learning, semi-supervised learning, active learning, and advanced clustering methods. The class is recommended for students with a deep interest in machine learning who have the appropriate prerequisites.
Prerequisites: Satisfactory performance in CMPSCI 689 as taught in Spring or Fall 2002, or permission of the instructors. These were the machine learning courses based on the book by Duda, Hart, and Stork. We will assume knowledge of Bayesian decision theory, maximum likelihood and Bayesian parameter estimation, perceptron and least-mean-square methods, basics of neural networks, basics of support vector machines, basics of computational learning theory, basics of clustering.
Credit: 3 units
Lecture: Tuesday & Thursday 2:30-3:45
Instructors:
Textbooks: The lectures will not follow any particular book, but there will be many assigned, suggested, and optional readings. We will put several books on reserve, but we recommend purchasing: Hastie, Tibshirani, Friedman: "The Elements of Statistical Learning", Springer.
Grading: Grading will be based on regular homework assignments, a course project, and class participation. Homework assignments will include both mathematical exercises and programming exercises. Late assignments will not be accepted once solutions have been distributed in class.