MATH 543 - Computational Statistics

Computation is at the heart of modern statistics and machine learning. This course is designed to acquire a comprehensive working knowledge of modern statistical computing. Topics include a review of numerical linear algebra, numerical optimization in statistical inference and machine learning (Newton-Raphson, expectation- maximization (EM) algorithm, Fisher scoring, stochastic gradient descent, etc.), a primer of convex optimization and applications (support vector machines, Gaussian processes, isotonic regression, LASSO, etc.), random number generation, Monte Carlo methods (sampling, simulations, MCMC, variational inference), randomization methods (jackknife, bootstrap, and permutation tests), estimation of functions (orthogonal polynomials, splines, etc.), density estimation. Offered regularly. Prerequisites: MATH 500 and MATH 530.