Inferring cosmic particles

par M. Rémi Bardenet (LAL, University Paris-Sud XI, IN2P3)

vendredi 13 avr. 2012, 11:00 → 12:00 Europe/Paris

Bat. 208, salle 100 (Universe)

INFERRING COSMIC PARTICLES
Challenges in MCMC methodology raised by the analysis of data from the Pierre Auger experiment.
Rémi Bardenet, LAL & LRI, Univ. Paris-Sud XI

I will start by presenting the Pierre Auger experiment, a giant ultra-high energy cosmic ray observatory located in Argentina, along with a simplified model of the signal it detects. An important part of my thesis has been dedicated to modelling this signal and work on novel inference methods to infer the parameters of the observed cosmic showers. After having presented the basics of Bayesian inference with MCMC techniques, I will concentrate on some of the challenges that naturally come up when performing Bayesian inference on the type of models encountered in Auger, and how we dealt with these challenges. Both the problems and the solution are generic enough to interest everyone interested in performing inference on the parameters of a signal made of several "bumps" or "components", the number of which is unknown.

Transparents slides.pdf