Statistical Learning Theory and Applications
Supervised learning from the perspective of modern statistical learning theory, taught at MIT. The course begins with multivariate function approximation from sparse data, then builds core tools including regularization and support vector machines for both regression and classification. Generalization bounds are derived using stability theory and VC theory, and the syllabus covers boosting and feature selection. Applications span computer vision, computer graphics, text classification, and bioinformatics, connecting the mathematics to practical systems. The course includes lecture materials, hands-on exercises, and a final project, giving students a chance to apply the techniques to real data. Offered through MIT OpenCourseWare, all materials are free to access under a Creative Commons license, though no certificate is awarded. Suited to students with a solid grounding in probability, linear algebra, and machine learning fundamentals.