Matrix Approximation under Local Low-Rank Assumption
Machine Learning
2013-01-16 v1 Machine Learning
Abstract
Matrix approximation is a common tool in machine learning for building accurate prediction models for 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 only 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 improvements in prediction accuracy in recommendation tasks.
Cite
@article{arxiv.1301.3192,
title = {Matrix Approximation under Local Low-Rank Assumption},
author = {Joonseok Lee and Seungyeon Kim and Guy Lebanon and Yoram Singer},
journal= {arXiv preprint arXiv:1301.3192},
year = {2013}
}
Comments
3 pages, 2 figures, Workshop submission to the First International Conference on Learning Representations (ICLR)