Speaker Guy Lebanon
Host Ran Gilad-Bachrach
Date recorded 6 May 2014
Matrix approximation is a common tool in recommendation systems, text mining, and computer vision. A prevalent assumption in constructing matrix approximations is that the partially observed matrix is of low-rank. We propose a new matrix approximation model where we assume instead that the matrix is locally of low-rank, leading to a representation of the observed matrix as a weighted sum of low-rank matrices. We analyze the accuracy of the proposed local low-rank modeling. Our experiments show significant improvements in prediction accuracy in the context of recommendation systems.
©2014 Microsoft Corporation. All rights reserved.