MA 576 Optimization models in Data science
This course will introduce the students to basic models and methods in machine learning and, more generally, in data-driven optimization and sequential decision making. The course will provide introduction to stochastic subgradient calculus for non-smooth convex functions, foundation of regularization in optimization problems and its role in numerical methods of optimization. Typical models of statistical estimation, such as various regression models, classification, and other problems in data science are analyzed as sample average approximation models. The second portion of the class discusses sequential decision models focusing specifically on their role in the process of statistical learning. Both Markov and Non-Markov finite-horizon decision problems will be introduced. Applications will cover decision trees, hierarchical clustering, Bayesian classifiers, and others as time permits. Some attention will be placed on bias- and variance reduction techniques.
Prerequisite
MA 540 and (Grad Student or (Junior or Senior))