BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Optimization in large dynamical systems: reducing the information 
 gap
DTSTART:20100624T151500
DTSTAMP:20260407T074518Z
UID:05006334b538a52bdf9b5925abbc3961c30b5c324b17aebcd09dad5b
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Bruno Gaujal\, INRIA\nThis talk  investigates the  behav
 ior and the control of discrete dynamic systems made of  distributed objec
 ts evolving in a common  environment.\nWe  build an information hierarchy\
 , from blind to clairvoyant information schemes on one hand\, and from dis
 tributed local information to central\, total information on the other. Th
 is hierarchy has a counterpart for the optimization procedures\, from stat
 ic  to adaptative policies on one side and from distributed selfish polici
 es to social optimal policies on the other.\n\nWe show that when the numbe
 r of objects  becomes large\, the optimal cost of the system converges to 
 the  optimal cost of a mean field limit system that is deterministic. Conv
 ergence also holds  for optimal policies. This implies that the value of i
 nformation disappears when the number of objects goes to infinity so that 
 the hierarchy constructed above collapses.\n\nThis framework is illustrate
 d by a brokering problem in grid  computing.\nWe provide several compariso
 ns in the blind versus clairvoyant cases as well as bounds on the price of
  anarchy (resulting from a lack of central information)  and show how this
  vanishes  when the size of the system grows.\n\nSeveral simulations with 
 growing numbers of processors will also be reported. They compare the perf
 ormance of the optimal policy of the limit system used in the finite case\
 ,  with classical policies by measuring its asymptotic gain. \n Prof. Gauj
 al's homepage 
LOCATION:BC 01 https://plan.epfl.ch/?room==BC%2001
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR