Probabilistic approaches to unsupervised learning
Course description:This course is a rigorous treatment of latent variable models and other probabilistic approaches to unsupervised learning. The models we'll cover include exponential families, mixture models, hidden Markov models, Dirichlet processes, topic models, and variational autoencoders. The class will emphasize algorithms with provable guarantees for sampling, inference, and learning.
Required knowledge: Familiarity with the basics of linear algebra, probability and statistics, and algorithms.