Recent advances in high dimensional robust statistics

It is classically understood how to learn the parameters of a Gaussian even in high dimensions from independent samples. However, estimators like the sample mean are very fragile to noise. In particular, a single corrupted sample can arbitrarily distort the sample mean. More generally we would like to be able to estimate the parameters of a distribution even if a small fraction of the samples are corrupted, potentially adversarially. Classically algorithms for this problem have all fallen into one of two categories: those whose error depends polynomially on the dimension (like the coordinate-wise median), and those whose runtimes are exponential in the dimension (like the Tukey median). We discuss recent work to overcome this barrier, achieving dimension independent error in polynomial time.

Date

Speakers

Daniel Kane

Affiliation

University of California, San Diego