← volver

Ciclo Seminarios Instituto Sistemas Complejos de Ingeniería (ISCI)


13:15 horas, Sala Asamblea, Beauchef Poniente, 4to piso (Beauchef 851, Santiago)

Invitado: Bruno Ziliotto, CNRS researcher in Mathematics, Paris Dauphine University

Título de la conferencia: “Partially Observable Markov Decision Processes with finite memory”, trabajo conjunto con K. Chatterjee y R. Saona.

A Partially Observable Markov Decision Process (POMDP) is a discrete-time repeated decision-problem where at each period, the stage payoff depends both on the stage action and on the current state of the world. The state evolves stochastically from one stage to the other. The decision-maker does not know the state, but receives a stream of signals about it. One example is a cleaning robot, that does not know the configuration of the room it is cleaning, but learns it while cleaning. We consider a long interaction, and prove that the decision-maker has approximately optimal strategies that have finite memory. This implies that the problem of computing approximately the long-term value of a POMDP is semi-decidable, which is surprising since it was known to be undecidable.

Almuerzos previa inscripción

Consultas a seminarios@isci.cl

Organiza: Instituto Sistemas Complejos de Ingeniería (ISCI)