Networks for Learning: Regression and Classification
MIT OpenCourseWare graduate course on supervised learning through Statistical Learning Theory. Coverage moves from classical regularization theory in reproducing kernel Hilbert spaces for function approximation to VC theory, which is used to justify methods such as Regularization Networks and Support Vector Machines. Additional topics include boosting, feature selection, and multiclass classification. Applications span computer vision, computer graphics, database search, and time series prediction, with a section connecting learning theory to neurobiology and object recognition in the brain. The course emphasizes hands-on exercises alongside the theoretical material. Materials include lecture notes, readings, and problem sets as published on MIT OpenCourseWare, free to access with no certificate offered.