“…Such compact representations which retain the key features of a high-dimensional matrix provide a significant reduction in memory requirements, and more importantly, computational costs when the latter scales, e.g., according to a high-degree polynomial, with the dimensionality. Matrices with low-rank structures have found many applications in background subtraction [1,2], system identification [3], IP network anomaly detection [4,5], latent variable graphical modeling [6], subspace clustering [7,8] and sensor and multichannel signal processing [9,10,11,12,13,14,15], [16,17,18,19,20,21,22,23,24,25,26,27]. .…”