Matrix Methods in Data Analysis, Signal Processing, and Machine Learning
Gilbert Strang's MIT course builds the linear algebra needed for modern data science, starting from matrix multiplication and factorization and moving through singular value decomposition, principal component analysis, and low-rank approximations. Lectures connect these tools to probability, statistics, and optimization, then apply them to regression, clustering, and neural networks, ending with a full treatment of deep learning architectures and backpropagation. Materials include video lectures, lecture notes, problem sets, and exams published through MIT OpenCourseWare, free to access with no certificate offered. The course assumes prior exposure to linear algebra and is aimed at students who want to see the matrix computations underneath machine learning rather than just the algorithms. Strang's own textbook on the subject parallels the syllabus, and the course is part of MIT's OCW math curriculum used in its 18.065 offering.