Robustness meets algorithms

In every corner of machine learning and statistics, there is a need for estimators that work not just in an idealized model, but even when their assumptions are violated. Unfortunately, high dimensions, being provably robust efficiently computable often at odds with each other. We give the first efficient algorithm estimating parameters high-dimensional Gaussian able to tolerate constant fraction corruptions independent dimension. Prior our work, all known either needed time exponential dimension compute or could only inverse-polynomial corruptions. Not does bridge gap between robustness algorithms, also it turns out be highly practical variety settings.