BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20190331T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20181028T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20190219T052539Z
UID:5c32fcfe5e3b9100996447@ist.ac.at
DTSTART:20190214T090000
DTEND:20190214T100000
DESCRIPTION:Speaker: Debarghya Ghoshdastidar\nhosted by Christoph Lampert\n
Abstract: Two-sample hypothesis testing for graphs is the statistical prob
lem of testing whether two given populations of graphs are similar or sign
ificantly different. It is an important tool in many scientific discipline
s including bioinformatics\, neuroscience and social science. For instance
\, testing between brain networks of Alzheimer patients and healthy indivi
duals reveal the neurological effects of Alzheimer's disease. The graph te
sting problem is quite challenging as one often needs to draw inference fr
om one or few samples of large graphs.

In this talk\, we provide insig
hts into the fundamental challenges of the problem from a statistical (min
imax) perspective. We show that some standard formulations of the testing
problem are unsolvable if we observe only few samples. On the positive sid
e\, we present two problem formulations that are solvable even when we obs
erve only one or two graphs from each population. We also present new stat
istical tests based on asymptotics of large random graphs\, and demonstrat
e the use of these methods in testing real networks.
LOCATION:Mondi Seminar Room 2\, Central Building\, IST Austria
ORGANIZER:tguggenb@ist.ac.at
SUMMARY:Hypothesis testing for graphs: Fundamental limits and practical met
hods
URL:https://talks-calendar.app.ist.ac.at/events/1727
END:VEVENT
END:VCALENDAR