We derive a convex relaxation for cardinality constrained Principal Component Analysis (PCA) by using a simple representation of the L1 unit ball and standard Lagrangian duality. ...
We analyze a class of estimators based on a convex relaxation for solving highdimensional matrix decomposition problems. The observations are the noisy realizations of the sum of ...
Alekh Agarwal, Sahand Negahban, Martin J. Wainwrig...