BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20180325T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20171029T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260424T142943Z
UID:5a27f0ecf0225726380189@ist.ac.at
DTSTART:20171215T094500
DTEND:20171215T104500
DESCRIPTION:Speaker: Yuri Boykov\nhosted by Vladimir Kolmogorov\nAbstract: 
 This talk discusses two seemingly unrelated data analysis methodologies: k
 ernel clustering and graphical models. Clustering is an unsupervised learn
 ing technique for generaldata where kernel methods are known for their dis
 criminating power. Graphical models such as Markov Random Fields (MRF) and
  related continuous geometric methods representcommon image segmentation m
 ethodologies. While both clustering and regularization models are very wid
 ely used in machine learning and computer vision\, they could not becombin
 ed before due to significant differences in the corresponding optimization
 \, e.g. spectral relaxation vs. combinatorial optimization methods. This t
 alk reviews thegeneral properties of kernel clustering and graphical model
 s\, discusses their limitations (including newly discovered "density biase
 s" in kernel methods)\, and proposes ageneral easy-to-implement algorithm 
 based on iterative bound optimization. In particular\, we show that popula
 r MRF potentials introduce principled geometric and contextualconstraints 
 into clustering\, while standard kernel methodology allows graphical model
 s to work with arbitrary high-dimensional features (e.g. RGBD\, RGBDXY\, d
 eep\, etc).
LOCATION:Mondi Seminar Room 3\, Central Building\, ISTA
ORGANIZER:abonvent@ist.ac.at
SUMMARY:Yuri Boykov: Kernel Clustering meets Graphical Models
URL:https://talks-calendar.ista.ac.at/events/992
END:VEVENT
END:VCALENDAR