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:CCIMI Seminars
SUMMARY:STORM: Stochastic Trust Region Framework with Rand
om Models - Katya Scheinberg\, Lehigh University
DTSTART;TZID=Europe/London:20170315T140000
DTEND;TZID=Europe/London:20170315T150000
UID:TALK71410AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/71410
DESCRIPTION:We will present a very general framework for uncon
strained stochastic optimization which is based on
standard trust region framework using random mod
els. In particular this framework retains the desi
rable features such step acceptance criterion\, tr
ust region adjustment and ability to utilize of se
cond order models. We make assumptions on the stoc
hasticity that are different from the typical assu
mptions of stochastic and simulation-based optimiz
ation. In particular we assume that our models and
function values satisfy some good quality conditi
ons with some probability fixed\, but can be arbit
rarily bad otherwise. We will analyze the converge
nce and convergence rates of this general framewor
k and discuss the requirement on the models and fu
nction values. We will will contrast our results w
ith existing results from stochastic approximation
literature. \nWe will motivate the framework with
examples of applications arising the area of mach
ine learning.
LOCATION:MR3 Centre for Mathematical Sciences
CONTACT:Rachel Furner
END:VEVENT
END:VCALENDAR