[Theory] TODAY: [TTIC Talks] 10/14 Research at TTIC: Baba C. Vemuri, University of Florida

[Brandie Jones]

*When:*         Friday, October 14th at *12:30pm 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=53793a5e-66c2-45cc-ac11-af2700e1817e>
)


*Who:*            Baba C. Vemuri, University of Florida



*Title:*            ManifoldNet: A Deep Neural Network for Manifold-valued
Data with Applications


*Abstract:    *Developing deep neural networks (DNNs) for manifold-valued
data sets has gained significant in terest of late in the deep learning
research community. Manifold-valued data abound many fields of Engineering
and Sciences including but not limited to, Medical Imaging, Computer
Vision, Robotics, etc., for example, diffusion tensor images (DTI), shape
(landmarks) data, directional data, covariance matrices, GPS data and
others. In this talk, a new theory and supporting architecture for DNNs
tailored for manifold-valued data inputs dubbed, ManifoldNet, will be
presented. Analogous to vector spaces where convolutions are equivalent to
computing weighted means, manifold-valued data convolutions will be defined
using the weighted Frechet Mean (wFM). To this end, a provably convergent
recursive ´ algorithm for computation of the wFM of the given data is
presented, where the weights are to be learned. Further, the proposed wFM
operator is provably equivariant to the natural group actions admitted by
the data manifold and achieves a contraction mapping. A novel network
architecture to realize the Mani foldNet will be detailed during the talk.
Experiments showcasing the performance of the ManifoldNet on regression and
classification problems in Neuroimaging will be presented. Finally, if time
permits, a generalization of the ManifoldNet to accommodate higher order
manifold-valued convolutions will be briefly discussed.




*Bio:     *Baba Vemuri received his PhD in Electrical and Computer
Engineering from the University of Texas at Austin in 1987. He then joined
the Department of Computer and Information Sciences
at the University of Florida, Gainesville, where he currently is a
Distinguished Professor and holds the Wilson and Marie Collins
Professorship in Engineering. His research interests include Geometric Deep
Learning, Geometric Statistics, Medical Image Computing, Computer Vision,
Machine Learning and Information Geometry. He has published over 200
refereed journal and conference articles in the aforementioned areas and
received several best paper awards. He has served as a program chair and
area chair of several IEEE sponsored conferences. He was an associate
editor for several area journals and is currently an associate editor for
the journal of Information Geometry. Professor Vemuri is a recipient of the
IEEE Computer Society’s Technical Achievement Award (2017) and is a fellow
of the IEEE (2001) and the ACM (2009).

***********************************************************************************************

*Presence at TTIC requires being fully vaccinated for COVID-19 or having
a TTIC or UChicago-approved exemption. Masks are optional in all common
areas. Full visitor guidance is available at ttic.edu/visitors
<http://ttic.edu/visitors>.*

***********************************************************************************************

*Research at TTIC Seminar Series*



TTIC is hosting a weekly seminar series presenting the research currently
underway at the Institute. Every week a different TTIC faculty member will
present their research.  The lectures are intended both for students
seeking research topics and advisors and for the general TTIC and
University of Chicago communities interested in hearing what their
colleagues are up to.




-- 
*Brandie Jones *
*Administrative Assistant*
Toyota Technological Institute
6045 S. Kenwood Avenue
Chicago, IL  60637
www.ttic.edu

Working Remote on Tuesdays
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/theory/attachments/20221014/7eda0606/attachment.html>