1: \begin{abstract}%   <- trailing '%' for backward compatibility of .sty file

2: 	Recent methods for learning a linear subspace from data corrupted by outliers are based on convex $\ell_1$ and nuclear norm optimization and require the dimension of the subspace and the number of outliers to be sufficiently small \citep{Xu:NIPS10}. In sharp contrast, the recently proposed \emph{Dual Principal Component Pursuit (DPCP)} method \citep{Tsakiris:DPCPICCV15} can provably handle subspaces of high dimension by solving a non-convex $\ell_1$ optimization problem on the sphere. However, its geometric analysis is based on quantities that are difficult to interpret and are not amenable to  statistical analysis. In this paper we provide a refined geometric analysis and a new statistical analysis that show that DPCP can tolerate as many outliers as the {\em square} of the number of inliers, thus improving upon other provably correct robust PCA methods. We also propose a scalable {\em Projected Sub-Gradient Method} method (DPCP-PSGM) for solving the DPCP problem and show it admits linear convergence even though the underlying optimization problem is non-convex and non-smooth. Experiments on road plane detection from 3D point cloud data demonstrate that DPCP-PSGM can be more efficient than the traditional RANSAC algorithm, which is one of the most popular methods for such computer vision applications.

3: 	\end{abstract}

4: