Thursday, May 5, 2022, 16:00
online only
(for the zoom link contact michael.spira@psi.ch, johannes.schlenk@psi.ch or
antonio.coutinho@psi.ch)
Giulio D'Agostini, University Sapienza Roma
Abstract:
Probabilistic reasoning is of greatest importance in tackling what
Poincaré used to call "the essential problem of the experimental
method", i.e. how to infer, from the observed effects, the causes that
have produced them, all times when there is no deterministic link
between causes (the parameters of our models of reality) and detectable
consequences (the experimental observations). The approach outlined is
basically that developed organically by Laplace, although it is
presently known with the appellative `Bayesian'.
The revival, in the last few decades, of probabilistic reasoning in
inference and forecasting is mainly due to unprecedentlty computing
power (together with several mathematical progresses), which finally
make possible to perform, although numerically or by Monte Carlo
methods, the required calculations. Particular emphasis will be given
to the conceptual and practical importance of the graphical
representation of the inferential and predictive problems ("Bayesian
networks").