BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Multiresolution Matrix Factorization
DTSTART:20161201T140000
DTEND:20161201T150000
DTSTAMP:20260427T215125Z
UID:b2716832602c6f51639b154bb645ab2bb33a01acfbaca7f332a4687a
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr Risi Kondor\, University of Chicago\nBio: Prof. Kondor's re
 search interests:\n\nMultiresolution/multiscale matrix factorizations\nMac
 hine learning for many-particle physics\nLearning graphs and other combina
 torial structures\nPermutation problems and Fourier analysis on the symmet
 ric group\nThe size of today's datasets dictates that machine learning alg
 orithms compress or reduce their input data and/or make use of parallelism
 . Multiresolution Matrix Factorization (MMF) makes a connection between su
 ch computational strategies and some classical themes in Applied Mathemati
 cs\, namely Multiresolution Analysis and Multigrid Methods. In particular\
 , the similarity (kernel) matrices appearing in data often have multiresol
 ution structure\, which can be exploited both for learning and to facilita
 te computation.\n\nMMF is an algorithm both for finding structure in large
  matrices (somewhat similar to HSS matrices)\, and constructing wavelet ba
 ses on graphs. I will highlight applications to matrix compression/sketchi
 ng and graph based semi-supervised learning. I will also present our paral
 lel MMF software library that allows the method to easily scale to sparse 
 matrices with ~10^6 rows/columns.\n\nThe work presented in this talk is jo
 int with my students Nedelina Teneva\, Pramod Mudrakarta\, Yi Ding and Vik
 as Garg.
LOCATION:ELA 1
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR