Robust polynomial regression up to the information theoretic limit

TL;DR

This paper introduces a new algorithm for robust polynomial regression that works efficiently for a wide range of outlier proportions, up to 50%, improving upon previous methods and establishing fundamental limits.

Contribution

The authors present a novel algorithm achieving robust polynomial regression for any outlier rate below 50%, surpassing prior constraints and providing tight impossibility bounds.

Findings

Algorithm works for outlier rate up to 1/2

Achieves a factor 2 approximation

Impossibility results show 1.09 approximation is impossible

Abstract

We consider the problem of robust polynomial regression, where one receives samples $(x_i, y_i)$ that are usually within $\sigma$ of a polynomial $y = p(x)$, but have a $\rho$ chance of being arbitrary adversarial outliers. Previously, it was known how to efficiently estimate $p$ only when $\rho < \frac{1}{\log d}$. We give an algorithm that works for the entire feasible range of $\rho < 1/2$, while simultaneously improving other parameters of the problem. We complement our algorithm, which gives a factor 2 approximation, with impossibility results that show, for example, that a $1.09$ approximation is impossible even with infinitely many samples.