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:CUED Control Group Seminars
SUMMARY:Gradient methods for huge-scale optimization probl
ems - Prof Yurii Nesterov\, Catholic University of
Louvain\, Belgium
DTSTART;TZID=Europe/London:20140527T173000
DTEND;TZID=Europe/London:20140527T183000
UID:TALK51798AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/51798
DESCRIPTION:We consider a new class of huge-scale problems\, t
he problems with sparse gradients. The most import
ant functions of this type are piece-wise linear.\
nFor optimization problems with uniform sparsity o
f corresponding linear operators\, we suggest a ve
ry efficient implementation of the iterations\, wh
ich total cost depends logarithmically in the dime
nsion. This technique is based on a recursive upda
te of the results of matrix/vector products and th
e values of symmetric functions. It works well\, f
or example\, for matrices with few nonzero diagona
ls and for max-type functions.\n\nWe show that the
updating technique can be efficiently coupled wit
h the simplest gradient methods. Similar results c
an be obtained for a new non- smooth random varian
t of a coordinate descent scheme. We present also
the promising results of preliminary computational
experiments and discuss extensions of this techni
que.\n\nBiography:\nYurii Nesterov is a professor
at the Catholic University of Louvain\, Belgium\,
where he is a member of the Center for Operations
Research and Econometrics (CORE). He is the autho
r of 4 monographs and more than 80 refereed papers
in leading optimization journals. He was awarded
with the Dantzig Prize 2000 given by SIAM and the
Mathematical Programming Society (for research hav
ing a major impact on the field of mathematical pr
ogramming)\, the John von Neumann Theory Prize 200
9 given by INFORMS\, the Charles Broyden prize 201
0 (for the best paper in Optimization Methods and
Software journal)\, and the Honorable Francqui Cha
ir (University of Liège\, 2011-2012). \nThe main d
irection of his research is the development of eff
icient numerical methods for convex and nonconvex
optimization problems supported by a global comple
xity analysis. The most important results are obta
ined for general interior-point methods (theory of
self-concordant functions)\, fast gradient method
s (smoothing technique)\, and global complexity an
alysis of the second-order schemes (cubic regulari
zation of the Newton's method).\n
LOCATION:Cambridge University Engineering Department\, LR6
CONTACT:Rachel Fogg
END:VEVENT
END:VCALENDAR