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:Signal Processing and Communications Lab Seminars
SUMMARY:On optimal sampling in off-the-grid sparse regular
isation. - Dr Clarice Poon\, DAMTP &\; Peterhou
se
DTSTART;TZID=Europe/London:20181129T150000
DTEND;TZID=Europe/London:20181129T160000
UID:TALK113926AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/113926
DESCRIPTION:Sparse regularization is a central technique for b
oth machine learning and imaging sciences. Existin
g performance guarantees assume a separation of th
e spikes based on an ad-hoc (usually Euclidean) mi
nimum distance condition\, which ignore the geomet
ry of the problem. In this talk\, we study the BLA
SSO (i.e. the off-the-grid version of L1 LASSO reg
ularization) and show that the Fisher-Rao distance
is the natural way to ensure and quantify support
recovery. Under a separation imposed by this dist
ance\, I will present results which show that stab
le recovery of a sparse measure can be achieved wh
en the sampling complexity is (up to log factors)
linear with sparsity. On deconvolution problems\,
which are translation invariant\, this generalizes
to the multi-dimensional setting\nexisting result
s of the literature. For more complex translation-
varying problems\, such as Laplace transform inver
sion\, this gives\nthe first geometry-aware guaran
tees for sparse recovery. This is joint work with
Nicolas Keriven and Gabriel Peyre.
LOCATION:LR12\, Baker Building\, CUED
CONTACT:Dr Ramji Venkataramanan
END:VEVENT
END:VCALENDAR