[Colloquium] [TTIC Talks] 1/27 Talks at TTIC: Theodor Misiakiewicz, Stanford University

[Brandie Jones]

*When: *Friday, January 27th* at 11 AM CT*

*Where: *Talk will be given *live, in-person* at

                   TTIC, 6045 S. Kenwood Avenue

                   5th Floor, Room 530


*Virtually: *via Panopto (Livestream
<https://uchicago.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=e737c92e-7907-49cc-8f72-af8e0114571d>
)

*Who: *Theodor Misiakiewicz, Stanford University

*Title: *New statistical and computational phenomena from Deep Learning

*Abstract:* Deep learning methodology has presented major challenges for
statistical learning theory. Indeed deep neural networks often operate in
regimes that are not explained or even contradict classical statistics and
optimization wisdom.  In this talk, we will consider two surprising
simulations illustrating some of these challenges. The first simulation
shows an instance where kernel ridge regression with a simple RBF kernel
achieves optimal test error when perfectly fitting the noisy training data.
Why can we interpolate noisy data and still generalize well? Why can
overfitting be benign in kernel ridge regression? In a second simulation,
we consider fitting two different smooth ridge functions with deep neural
networks (DNNs). Both can be estimated at the same near-parametric rate by
DNNs, if we have unbounded computational resources. However, our simulation
shows that learning becomes much harder for one of these functions when
restricted to DNNs trained using SGD. Why does SGD succeed on some
functions and fail on others? The goal of this talk will be to understand
these two simulations. In particular, we will show that we can develop
quantitative theories that can precisely capture both phenomena.

*Bio: *Theodor is a sixth-year PhD candidate in Statistics at Stanford
University working with Prof. Andrea Montanari. Before his Ph.D, he
completed his undergraduate studies in France at Ecole Normale Superieure
de Paris with a focus on pure math and physics. He received a B.Sc. in
Mathematics and a B.Sc. in Physics in 2014, and a M.Sc. in Theoretical
Physics at ICFP in 2016. His interest lies broadly at the intersection of
statistics, machine learning, probability, and computer science.

For more information about him and his work, please visit his personal
webpage. <https://misiakie.github.io/>

*Host: Nati Srebro <nati at ttic.edu>*



-- 
*Brandie Jones *
*Executive **Administrative Assistant*
Toyota Technological Institute
6045 S. Kenwood Avenue
Chicago, IL  60637
www.ttic.edu
Working Remotely on Tuesdays
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20230120/9525c3ac/attachment-0001.html>