L. Belasko Garate, J. Doncel
We analyze a variant of the SIR model where susceptible population can avoid to get the infection and become recovered with a vaccine. We consider that the susceptible population can be protected from the epidemic following some confinement policy. We assume that the infected population incurs a cost and that the cost of confinement decreases with the susceptible population's exposure level.
We aim to analyze the optimal confinement policy in this context. We first model this problem as a Markov Decision Process with infinite horizon and discounted cost and we show that the optimal confinement strategy is of threshold-type. We also tackle this problem from the perspective of Reinforcement Learning, where no information about the dynamics of the systems and the costs of the population is known.
Palabras clave: Markov Decision Process, SIR model, Reinforcement Learning
Programado
Procesos Estocásticos y sus aplicaciones I
13 de junio de 2025 11:00
Auditorio 2. Leandre Cristòfol