Stat 314 – Mathematical problems in Machine Learning

This class will be an introduction to mathematical techniques to establish sharp asymptotics in high-dimensional statistics. Topics include:

• Universality.

• Asymptotics for convex empirical risk minimization and M estimation.

• Gaussian comparison inequalities.

• Approximate message passing algorithms.

• Sharp high-dimensional asymptotics of general first order algorithms.

• Bayes error in high-dimensional estimation.

• Non-convex problems.

Left: Comparison of two phase retrieval algorithms (column: algorithm | row: noise level)

Class Times and Location

• Tue-Thu, 9:00-10:20AM

• Room 200-305