Simultaneous Robust Matching Pursuit for Multi-view Learning