Algorithms for Inference
MIT OpenCourseWare offers this graduate-level introduction to statistical inference using probabilistic graphical models. The course covers Bayesian and hidden Markov models, belief propagation, the sum-product algorithm, Kalman filtering, and Monte Carlo methods, building the mathematical foundation shared by machine learning, signal processing, computer vision, and control systems. Materials include lecture notes, problem sets, and exams from MIT's Electrical Engineering and Computer Science department, taught at the level expected of graduate students with a background in probability. The course builds from exact inference in tree-structured graphs toward approximate inference methods needed for the large, loopy models common in real applications. As with other MIT OCW offerings, all lecture materials are free to access under a Creative Commons license, with no certificate offered since this is a self-study archive rather than an instructor-led session.