Jean Barbier's online seminar: "Information-theoretic limits in high-dimensional Bayesian inference: Can the signal be extracted from the noise?"

[Antonio Lerario]

ONLINE SEMINAR ANNOUNCEMENT

Speaker: Jean Barbier (ICTP)
Title: Information-theoretic limits in high-dimensional Bayesian inference:
Can the signal be extracted from the noise?

Venue: Apr 14, 2020 04:00 PM Rome
Zoom link: https://us04web.zoom.us/j/74258369797
Meeting ID: 742 5836 9797

Note: the seminar will consists of two parts: an introductory part from
04:00 PM to 04:50 PM, followed by a little break, then a second part from
05:00 PM to 05:30 PM with questions and/or more advanced topics.

Abstract: I will emphasize on the deep links between physics and Bayesian
inference through the description of some paradigmatic models in
signal-processing and machine learning thanks to the language of
statistical mechanics. I will then review a number of recent analytical
tools, in particular interpolation techniques, that allow to answer one of
the most fundamental information-theoretic question: given a model of the
form data(signal) + noise, when does the signal can be theoretically
extracted as a function of the noise level, or the amount of data? This
question is at the core of communications, for which information theory has
been developed, but is also key in the understanding of basic models in
signal processing and learning.Everyone is much welcome to join!

Everyone is much welcome to join!

Best,
Antonio

-- 
*http://people.sissa.it/~lerario <http://people.sissa.it/~lerario>*