This website summarizes the notes of Stanford's introductory course on probabilistic graphical models. It starts from the very basics and concludes by explaining from first principles the variational auto-encoder, an important probabilistic model that is also one of the most influential recent results in deep learning.