CSE 291: Robust Statistics

Fall 2023








Course Description: Most standard statistical estimators fail badly when faced with even a few extreme outliers. While naive outlier removal techniques can often solve this problem in low dimensions, the performance of such techniques degrades badly for higher dimensional problems. Until recently all known techniques for solving even simple statistical problems either produced errors scaling polynomially in the dimension or runtimes that were exponential in it. CSE 291 covers some recent breakthroughs in the field of computational statistics for the first time leading to practical solutions to many of these problems.