Search this site:

Directors:
Randolph E. Bank
Philip E. Gill
Michael Holst

Terry Le

Office: AP&M 7431
Phone: (858)534-9813
Fax: (858)534-5273
E-mail: tele@ucsd.edu
An overview of low-rank matrix recovery from incomplete measurement

Jingwen Liang
UCSD

Abstract:

The low-rank matrix recovery problem consists of reconstructing an unknown low-rank matrix (say a rank $r$ matrix in $\mathbb{R}^{n_1 \times n_2}$) from $m$ ($r \le m \ll \min\{n_1,n_2\}$) linear measurements. In this talk, we cover some recent results in low-rank matrix recovery. Specific attention is paid to the algorithm most commonly used in practice($\ell_1$ minimization) and the reconstruction guarantees that hold with high probability for these algorithms. We also covere the recent result on weighted $\ell_1$ minimization using the estimate of column and row space of the target matrix.

Tuesday, February 7, 2017
11:00AM AP&M 2402