BEGIN:VCALENDAR
PRODID:-//AT Content Types//AT Event//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20181011T113000Z
DTEND:20181011T123000Z
DCREATED:20180920T093201Z
UID:ATEvent-fec5fd6547604711b1e6d0870aaa010f
SEQUENCE:0
LAST-MODIFIED:20180920T124539Z
SUMMARY:Talk by Andreas Loukas (EPFL) : Graph reduction by local variation
DESCRIPTION:Can we reduce the size of a graph without significantly al
tering its basic properties? \nWe will approach the graph reduction pr
oblem from the perspective of restricted similarity\, \na modification
of a well-known measure for graph approximation. Our choice is motiva
ted \nby the observation that restricted similarity implies strong spe
ctral guarantees and can be \nused to prove statements about certain u
nsupervised learning problems. The talk will then \nfocus on coarsenin
g\, a popular type of graph reduction. We will derive sufficient condi
tions \nfor a small graph to approximate a larger one in the sense of
restricted similarity. Our findings \ngive rise to nearly-linear coars
ening algorithms that find coarse graphs of improved quality,\noften b
y a large margin\, without sacrificing speed.
LOCATION:M7.101
PRIORITY:3
TRANSP:0
END:VEVENT
END:VCALENDAR