6eme Journée COSMOS

Onglets principaux

Organisateur:

Intitulé: 
6eme Journées COSMOS avec ouverture sur la modélisation Markovienne
Date: 
Mercredi, 21 Juin, 2017
Formulaire d'inscription (voir onglet Register ci-dessus): 

 
Les 6ème journées COSMOS ont lieu en partenariat avec l'ANR Marmote.

Place
Salle 405 tour  24-25.
LIP6, Université Pierre et Marie Curie
4 place Jussieu Paris
How to go

Registration
Here or in the register tab above.

Speakers

  • Michel de Lara Rationally Biased Learning
  • Balakrishna Prahbu : Asymptotics of Insensitive Load Balancing and Blocking Phases.
  • Vincent Danos Macroeconomics of the cell: a simple predictive model of stochastic cell growth (joint work with Guillaume Terradot, Philipp Thomas and Andrea Weisse)
  • Alexey Piunovskiy Transformation of continuous-time MDP to the discrete-time model: case of the total expected cost.
  • Floske Spieksma : Continuity properties of parametrised and controlled countable state Markov processes.
  • Benjamin Legros A Uniformization Approach for the Dynamic Control of Queueing Systems with Abandonments (joint work with O. Jouini and G. Koole)
  • Alain Jean-Marie Optimal control of service in a queue with impatience and setup costs (joint work with E. Hyon)

Abstracts

  • Michel de Lara
    Rationally Biased Learning
    Slides
  • Balakrishna Prahbu
    Asymptotics of Insensitive Load Balancing and Blocking Phases
    Résumé : Load balancing with various type of load information has become a key component of modern communication and information systems. In many systems, characterizing precisely the blocking probability allows to establish a performance tradeoff between delay and losses. We address here the problem of giving robust performance bounds based on the study of the asymptotic behavior of the so-called insensitive load balancing schemes when the number of servers and the load scales jointly. These schemes have the desirable property that the stationary distribution of the resulting stochastic network depends on the distribution of job sizes only through its mean. It was shown that they give good estimates of performance indicators for systems with finite buffers, generalizing henceforth Erlang’s formula whereas optimal policies are already theoretically and computationally out of reach for networks of moderate size. We study the case of a single class of traffic acting on a set of processor sharing or symmetric queues with finite buffers, with load scaling with the number of servers. We characterize, using law of large numbers, central limit theorems and large deviations, the response of symmetric systems under those schemes. In particular, we show that there is a phase transition for the blocking probability. Before a critical load, the blocking is exponentially small in the number of the servers and becomes of the order of the inverse of the square root beyond this critical load. This generalizes the well-known Jagerman-Halfin-Whitt regime for a one-dimensional queue, and gives a generalized staffing rule for a given target blocking probability.
    Paper:    Here
  • Floske Spieksma :
    Continuity properties of parametrised and controlled countable state Markov processes
    Paper Chapter : Structures of optimal policies in Markov Decision Processes with unbounded jumps: the State of our Art in book Markov decision Processes in Practice.
    Slides