Theoretical Machine Learning

The course covers fundamental results in statistical learning theory: 1. Supervised learning: generalization, uniform concentration, empirical risk minimizer, Rademacher complexity, VC theory, reproducing Hilbert kernel space and several applications including neural networks, sparse linear regression, and low-rank matrix problems; 2. Online learning: sequential Rademacher complexity, Littlestone dimension, online algorithms and applications; 3. Unsupervised learning: latent variable models, maximum likelihood estimation, method of moments, tensor methods.