Algorithmic Aspects of Machine Learning

Catalog description

In this course, we will explore the theoretical foundations of machine learning and deep learning. We will focus on designing and analyzing machine learning algorithms with provable guarantees. More specifically, in this course we will (1) introduce basic tools in linear algebra and optimization, including the power method, singular value decomposition, matrix calculus, (matrix) concentration inequalities, and (stochastic) gradient descent, (2) cover many examples where one can design algorithms with provably guarantees for fundamental problems in machine learning (under certain assumptions), including topic modeling, tensor decomposition, sparse coding, and matrix completion, and (3) discuss the emerging theory of deep learning, including landscape analysis, generalization and over-parameterization, neural tangent kernels, generalization bounds, and implicit regularization.