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:Statistics
SUMMARY:On the Computational and Statistical Interface and
"\;Big Data"\; - Michael I. Jordan\, Univ
ersity of California\, Berkeley
DTSTART;TZID=Europe/London:20130510T154500
DTEND;TZID=Europe/London:20130510T164500
UID:TALK45237AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/45237
DESCRIPTION:The rapid growth in the size and scope of datasets
in science and technology\nhas\ncreated a need fo
r novel foundational perspectives on data analysis
that\nblend\nthe statistical and computational sc
iences. That classical perspectives\nfrom these f
ields are not adequate to address emerging problem
s in "Big\nData" is apparent\nfrom their sharply d
ivergent nature at an elementary level---in comput
er\nscience\, the growth of the number of data poi
nts is a source of "complexity"\nthat must be tame
d via algorithms or hardware\, whereas in statisti
cs\, the\ngrowth\nof the number of data points is
a source of "simplicity" in that inferences\nare g
enerally stronger and asymptotic results can be in
voked. Indeed\, if\ndata are a statistician's pri
ncipal resource\, why should more data be\nburdens
ome\nin some sense? Shouldn't it be possible to e
xploit the increasing\ninferential strength of dat
a at scale to keep computational complexity at\nba
y? I present three research vignettes that pursue
this theme\, the first\ninvolving the deployment
of resampling methods such as the bootstrap on\npa
rallel and distributed computing platforms\, the s
econd involving\nlarge-scale matrix completion\, a
nd the third introducing a methodology of\n"algori
thmic weakening\," whereby hierarchies of convex r
elaxations are used\nto control statistical risk a
s data accrue.\n\nJoint work with Venkat\nChandras
ekaran\, Ariel Kleiner\, Lester Mackey\, Purna Sar
kar\, and Ameet\nTalwalkar
LOCATION:Seminar Room 1\, Isaac Newton Institute
CONTACT:Richard Samworth
END:VEVENT
END:VCALENDAR