Deep Learning
MIT OpenCourseWare offers this intermediate course covering the theory and practice of deep learning. Topics include neural network architectures such as multilayer perceptrons, convolutional networks, recurrent networks, graph neural networks, and transformers, along with the geometry and invariances underlying them. The course works through backpropagation and automatic differentiation as the mechanics of training, then moves into learning theory and generalization in high dimensional settings. Applications span computer vision, natural language processing, and robotics, connecting the mathematical foundations to working systems. Materials come from MIT's Electrical Engineering and Computer Science department and follow the OpenCourseWare format of lecture notes, problem sets, and readings, released under a Creative Commons license for self-study at no cost.