1: \begin{abstract}

2: Despite its impressive theory \& practical performance, Frequent Directions (\acrshort{fd})

3: has not been widely adopted for large-scale regression tasks.

4: Prior work has shown randomized sketches 

5: (i) perform worse in estimating the covariance matrix of the data than \acrshort{fd};

6: (ii) incur high error when estimating the bias and/or variance on sketched ridge regression.

7: We give the first constant factor relative error bounds on the bias \& variance for sketched ridge 

8: regression using \acrshort{fd}.

9: We complement these statistical results by showing that \acrshort{fd} can be used in 

10: the optimization setting through an iterative

11: scheme which yields high-accuracy solutions.

12: This improves on randomized approaches which need to compromise the need for a new sketch every 

13: iteration with speed of convergence.

14: In both settings, we also show using \emph{Robust Frequent Directions} 

15: further enhances performance.

16: \end{abstract}

17: