BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:STORM: Stochastic Trust Region Framework with Random Models - Katy
 a Scheinberg\, Lehigh University
DTSTART:20170315T140000Z
DTEND:20170315T150000Z
UID:TALK71410@talks.cam.ac.uk
CONTACT:Rachel Furner
DESCRIPTION:We will present a very general framework for unconstrained sto
 chastic optimization which is based on standard trust region framework usi
 ng  random models. In particular this framework retains the desirable feat
 ures such step acceptance criterion\, trust region adjustment and ability 
 to utilize of second order models. We make assumptions on the stochasticit
 y that are different from the typical assumptions of stochastic and simula
 tion-based optimization. In particular we assume that our models and funct
 ion values satisfy some good quality conditions with some probability fixe
 d\, but can be arbitrarily bad otherwise. We will analyze the convergence 
 and convergence rates of this general framework and discuss the requiremen
 t on the models and function values. We will will contrast our results wit
 h existing results from stochastic approximation literature. \nWe will mot
 ivate the framework with examples of applications arising the area of mach
 ine learning. 
LOCATION:MR3 Centre for Mathematical Sciences
END:VEVENT
END:VCALENDAR