• This event has passed.

# Statistics Seminar: Jianqing Fan, PhD, Princeton University

## Jianqing Fan, Ph.D.

Professor of Statistics
Department of Operations Research
and Financial Engineering

### “Robust Low-Rank Matrix Recovery”

Abstract: This paper focuses on robust estimation of the low-rank matrix from the trace regression model. It encompasses four popularly problems: Sparse linear models, compressive sensing, matrix completion and multi-task regression as its specific examples. Instead of optimizing nuclear-norm penalized least-squares, our robust penalized leastsquares approach is to replace the quadratic loss by its robust version. The robust version is obtained by the appropriate truncations or shrinkage of the data and hence is very easy to implement. Under only bounded $2+\delta$ moment condition on the noise, we show that the proposed robust penalized trace regression yields an estimator that processes the same rates as those presented in Negahban and Wainwright’s work under sub-Gaussian error assumption. The rates of convergence are explicitly derived. As a byproduct, we also give a robust covariance matrix estimation and establish its concentration inequality.

Reception immediately following in room 450, West Hall (Lounge).

### Details

Date:
September 8, 2016
Time:
1:00 pm - 2:00 pm

## Venue

Room 411 West Hall
1085 S University Ave
Ann Arbor, MI 48109 United States
+ Google Map