BEGIN:VCALENDAR
CALSCALE:GREGORIAN
PRODID:iCalendar-Ruby
VERSION:2.0
BEGIN:VEVENT
DESCRIPTION: The method of moments is a statistical technique for density e
stimation that solves a system of moment equations to estimate the paramete
rs of an unknown distribution. A fundamental question critical to understan
ding identifiability asks how many moment equations are needed to get finit
ely many solutions and how many solutions there are. Since the moments o
f a mixture of Gaussians are polynomial expressions in the means\, variance
s and mixture weights\, one can address this question from the perspective
of algebraic geometry. With the help of tools from polyhedral geometry\, we
answer this fundamental question for several classes of Gaussian mixture m
odels. Furthermore\, these results allow us to present an algorithm that pe
rforms parameter recovery and density estimation\, applicable even in the h
igh dimensional case. Based on joint work with Julia Lindberg and Jose Ro
driguez (University of Wisconsin-Madison).
DTSTAMP:20220208T164600
DTSTART:20220214T141500
CLASS:PUBLIC
LOCATION:Online via Zoom.
SEQUENCE:0
SUMMARY:Carlos Amendola (Technische Universität München): Estimating Gaussi
an mixtures using sparse polynomial moment systems
UID:107919926@www.facetsofcomplexity.de
URL:http://www.facetsofcomplexity.de/monday/20220214-L-Amendola.html
END:VEVENT
END:VCALENDAR