[Colloquium] Teneva/Dissertation Defense/Jul 14, 2017

[Margaret Jaffey via Colloquium]

       Department of Computer Science/The University of Chicago

                     *** Dissertation Defense ***


Candidate:  Nedelina Teneva

Date:  Friday, July 14, 2017

Time:  12:00 PM

Place:  Ryerson 277

Title: Multiresolution matrix factorization

Abstract:
In this thesis we introduce a new type of structure in matrices,
called multiresolution factor- izability, which is an alternative to
the ubiquitous low rank assumption in machine learning and numerical
linear algebra. We show the connections between classical Fourier,
wavelet and multiresolution analysis – three concepts which have
shaped most of applied math in the last couple of decades – and (low
rank) matrix factorizations, which are often implicitly present in
machine learning algorithms as subroutines and which have been one of
the main drivers of the scalability of these algorithms in recent
years. We propose several different Multiresolution Matrix
Factorization (MMF) algorithms, some of which, like Parallel MMF
(publicly available in the form of a C++ software library) scale to
modern size data, and demonstrate that MMF can be used for compression
of the type of large scale matrices and graphs which typically arise
in machine learning applications.

Nedelina's advisor is Prof. Risi Kondor

Login to the Computer Science Department website for details,
including a draft copy of the dissertation:

 https://www.cs.uchicago.edu/phd/phd_announcements#nteneva

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Margaret P. Jaffey            margaret at cs.uchicago.edu
Department of Computer Science
Student Support Rep (Ry 156)               (773) 702-6011
The University of Chicago      http://www.cs.uchicago.edu
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=