BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Isaac Newton Institute Seminar Series
SUMMARY:Layered\, error enhanced hierarchical dictionary l
 earning algorithm for sparse coding. - Daniela Cal
 vetti (Case Western Reserve University)
DTSTART;TZID=Europe/London:20230428T113000
DTEND;TZID=Europe/London:20230428T123000
UID:TALK198454AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/198454
DESCRIPTION:In this talk we preset a novel multi-phase diction
 ary learning algorithm that addresses the complexi
 ty by clustering and reducing the dictionary\, enh
 ancing the resolution power of the method by accou
 nting for the representation error introduced by t
 he dictionary reduction. The problem is set up and
  solved in the Bayesian framework\, and all steps 
 involving sparse coding are performed by using spa
 rsity promoting Bayesian hypermodels and a priorco
 nditioning techniques that are demonstrated earlie
 r to provide a computationally efficient way to fi
 nd compressible solutions to linear inverse proble
 ms. As a novelty\, in the cluster identification p
 roblem\, we introduce a new and data-informed way 
 to implement group sparsity in order to identify a
 s few clusters as possible to explain the data. Mo
 reover\, ideas from the previous works on Bayesian
  modeling error analysis are modefied and extended
  to quantify the modeling error introduced when pa
 ssing from the full dictionary cluster to the redu
 ced one.&nbsp\;\nThis works has been done in colla
 boration with Alberto Bocchinfuso and Erkki Somers
 alo.\n&nbsp\;
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:
END:VEVENT
END:VCALENDAR