A brief introduction to Sequential Monte Carlo

schedule le lundi 22 octobre 2018 de 17h00 à 18h00

Organisé par : B. Dembin, G. Conchon-Kerjan, F. Coppini, O. Safsafi

Intervenant : Qiming Du (LPSM)
Lieu : Jussieu, salle Paul Lévy, couloir 16-26, salle 113

Sujet : A brief introduction to Sequential Monte Carlo

Résumé :

Sequential Monte Carlo (SMC) is a popular genetic type selection-mutation particle algorithm, widely used in signal processing, Bayesian statistical inference and rare event simulation. The more general framework is also called Feynman-Kac particle method. Starting with a brief introduction to the famous Feynman-Kac formula, I will try to explain the connection between the Feynman-Kac path measures and its particle approximation in the discret time settings, in order to give a better intuition of the natural ideas behind the construction of SMC method. Some typical applications such as nonlinear filtering in Hidden Markov Models (Particle Filters), rare event simulation in molecular dynamics (Multilevel Splitting) will also be provided.