Optimization for Data Science

Catalog description

The class presents the theory and algorithms for linear and nonlinear optimization problems with continuous variables. Topics covered include convex analysis, first- and second-order optimality methods, duality, KKT conditions, algorithms for unconstrained optimization, linearly and nonlinearly constrained problems, and convergence rates of algorithms. The theory and algorithms are applied to data science problems arising in machine learning, statistics, and related fields (e.g., maximizing likelihood and penalized likelihood functions, MM algorithms, EM algorithms, model selection, Sparse PCA). Students are expected to be comfortable with rigorous mathematical arguments.