Seminar: "Optimization of large Markov dynamic systems, with several applications" - Prof. Bruno Gaujal (INRIA researcher)

Seminar organized within the project DISMA Excellence Project 2018-2022.

"Optimization of large Markov dynamic systems, with several applications"

Prof. Bruno Gaujal (INRIA research director, Grenoble)

Tuesday January 25, 2021 at 16:30

Webinar: link

Abstract: In this talk I will explain how to deal with Markov decision processes, composed of a large number of objects and avoid the curse of dimensionality. I will show that that the optimal value of such a Markov decision process, which satisfies a Bellman equation, converges to the solution of a continuous Hamilton-Jacobi-Bellman (HJB) equation based on the mean field approximation of the Markov decision process. I will also give bounds on the difference of their values and present an algorithm for deriving an approximating solution to the Markov decision process from a solution of the HJB equations. I will illustrate the method on several examples such as investment strategies, population dynamics control and scheduling in queues. I will also mention pitfalls of this approach as well as extensions to systems with no drifts and systems with ab sorbing states.

Flyer of the event: the flyer of the event can be downloaded here