[Colloquium] Re: REMINDER: 8/22 TTIC Colloquium: Yudong Chen, Cornell

[Mary Marre via Colloquium]

When:     Monday, August 22nd at 11:00 a.m.

Where:    TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526

Who:      Yudong Chen, Cornell


Title:      Robustness and Flexibility in Non-convex Low-rank Estimation

Abstract:
Many problems in statistics and machine learning involve fitting a low-rank
matrix to noisy data. Recently developed non-convex methods, which work
directly with low-rank factorizations, provide an attractive computational
solution that overcomes the scalability limitations of the popular convex
relaxation approach. A natural question arises: can non-convex methods
match the well-documented statistical power of their convex counterpart?

In this talk, we provide insights to this question by demonstrating the
robustness and flexibility of non-convex methods. We robustify non-convex
methods to protect against arbitrarily corrupted data points, achieving the
same robustness guarantees as convexified robust PCA algorithms. We further
show that non-convex methods apply to settings with noise, constraints,
nonlinear measurements and non-quadratic loss functions, which cover a
broad range of problems including one-bit matrix completion, Sparse PCA,
community detection and matrix decomposition. For all these problems we
provide rigorous convergence guarantees in the presence of non-convexity,
and obtain near-optimal statistical bounds for the resulting solutions.
Computationally, we show that non-convex methods have overall running time
that is often near-linear.


Biography:
Yudong Chen is an assistant professor at the School of Operations Research
and Information Engineering (ORIE), Cornell University. In 2013-2015 he was
a postdoctoral scholar at the Department of Electrical Engineering and
Computer Sciences at University of California, Berkeley. He obtained his
Ph.D. in Electrical and Computer Engineering from the University of Texas
at Austin in 2013, and his M.S. and B.S. from Tsinghua University. His
research interests include machine learning, high-dimensional and robust
statistics, convex and non-convex optimization, and applications in
networks and financial systems.


Host: Srinadh Bhojanapalli, srinadh at ttic.edu



For more information on the colloquium series or to subscribe to the
mailing list, please see http://www.ttic.edu/colloquium.php


Mary C. Marre
Administrative Assistant
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 504*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*

On Sun, Aug 21, 2016 at 7:19 PM, Mary Marre <mmarre at ttic.edu> wrote:

> When:     Monday, August 22nd at 11:00 a.m.
>
> Where:    TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526
>
> Who:      Yudong Chen, Cornell
>
>
> Title:      Robustness and Flexibility in Non-convex Low-rank Estimation
>
> Abstract:
> Many problems in statistics and machine learning involve fitting a
> low-rank matrix to noisy data. Recently developed non-convex methods, which
> work directly with low-rank factorizations, provide an attractive
> computational solution that overcomes the scalability limitations of the
> popular convex relaxation approach. A natural question arises: can
> non-convex methods match the well-documented statistical power of their
> convex counterpart?
>
> In this talk, we provide insights to this question by demonstrating the
> robustness and flexibility of non-convex methods. We robustify non-convex
> methods to protect against arbitrarily corrupted data points, achieving the
> same robustness guarantees as convexified robust PCA algorithms. We further
> show that non-convex methods apply to settings with noise, constraints,
> nonlinear measurements and non-quadratic loss functions, which cover a
> broad range of problems including one-bit matrix completion, Sparse PCA,
> community detection and matrix decomposition. For all these problems we
> provide rigorous convergence guarantees in the presence of non-convexity,
> and obtain near-optimal statistical bounds for the resulting solutions.
> Computationally, we show that non-convex methods have overall running time
> that is often near-linear.
>
>
> Biography:
> Yudong Chen is an assistant professor at the School of Operations Research
> and Information Engineering (ORIE), Cornell University. In 2013-2015 he was
> a postdoctoral scholar at the Department of Electrical Engineering and
> Computer Sciences at University of California, Berkeley. He obtained his
> Ph.D. in Electrical and Computer Engineering from the University of Texas
> at Austin in 2013, and his M.S. and B.S. from Tsinghua University. His
> research interests include machine learning, high-dimensional and robust
> statistics, convex and non-convex optimization, and applications in
> networks and financial systems.
>
>
> Host: Srinadh Bhojanapalli, srinadh at ttic.edu
>
>
>
> For more information on the colloquium series or to subscribe to the
> mailing list, please see http://www.ttic.edu/colloquium.php
>
>
>
> Mary C. Marre
> Administrative Assistant
> *Toyota Technological Institute*
> *6045 S. Kenwood Avenue*
> *Room 504*
> *Chicago, IL  60637*
> *p:(773) 834-1757 <%28773%29%20834-1757>*
> *f: (773) 357-6970 <%28773%29%20357-6970>*
> *mmarre at ttic.edu <mmarre at ttic.edu>*
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20160822/36ae3419/attachment-0001.html>