High-dimensional deformed rectangular matrices and noise reduction
Speaker:
Xiucai Ding, University of Toronto
Date and Time:
Friday, February 10, 2017 - 2:10pm to 3:00pm
Location:
Fields Institute, Stewart Library
Abstract:
We consider the recovery of a low-rank matrix from its noisy observation under two regimes. To construct the optimal estimators, we compute the convergent limit and optimal rate for the singular values and vectors.
The main tools we use are the isotropic local Marchenko-Pastur law and a linearization technique.