Nips 2011 Sparse Representation Low Rank Approximation Workshop Fast Memory Google Techtalks