Topics that are covered in the course include: " Bayesian decision theory: the theoretical statistical basis for recognition based on Bayes theorem from probability " Maximum-likelihood and Bayesian parameter estimation: parameters of probability density functions " Nonparametric techniques: Parzen window, k-nearest neighbor " Linear discriminant functions: gradient descent, relaxation, minimum squared-error procedures such as LMS, and support vector machines " Algorithm-independent machine learning " Unsupervised learning and clustering The course is quite mathematical. Students enrolling this class are expected to have a good understanding of probability and random variables, both one-dimensional and multi-dimensional, and a good background in linear algebra as well as calculus. Some of the necessary math will be reviewed at the beginning of the course, but it is only a quick review, not a math course. Grades will be based on homework, tests, and small computer projects. |