Optimization in large dynamical systems: reducing the information gap
Event details
| Date | 24.06.2010 |
| Hour | 15:15 |
| Speaker | Prof. Bruno Gaujal, INRIA |
| Location | |
| Category | Conferences - Seminars |
This talk investigates the behavior and the control of discrete dynamic systems made of distributed objects evolving in a common environment.
We build an information hierarchy, from blind to clairvoyant information schemes on one hand, and from distributed local information to central, total information on the other. This hierarchy has a counterpart for the optimization procedures, from static to adaptative policies on one side and from distributed selfish policies to social optimal policies on the other.
We show that when the number of objects becomes large, the optimal cost of the system converges to the optimal cost of a mean field limit system that is deterministic. Convergence also holds for optimal policies. This implies that the value of information disappears when the number of objects goes to infinity so that the hierarchy constructed above collapses.
This framework is illustrated by a brokering problem in grid computing.
We provide several comparisons 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.
Several simulations with growing numbers of processors will also be reported. They compare the performance of the optimal policy of the limit system used in the finite case, with classical policies by measuring its asymptotic gain.
Prof. Gaujal's homepage
Practical information
- General public
- Free