Andrea Montanari
L'Ecole Normale Superieure, Paris
Thursday, February 16, 2006
4:15pm
Packard 101
Abstract:
A central theme in many fields of science (from probability to statistical mechanics and system biology) is to understand the collective behavior of large assemblies of elementary components, each one interacting with a few neighbors. The same topic is increasingly relevant to engineering, computer science and statistics, giving rise to a whole new field of research at the crossroad of several disciplines.
I will discuss three specific examples from my own work: (i) Multi-user detection; (ii) Random constraint satisfaction problems; (iii) Modern (iterative/probabilistic) coding systems (e.g. Turbo or LDPC codes). Special attention will be devoted to the last topic.
In these examples, an initial non-rigorous calculation based on physical ideas, can be turned into a rigorous mathematical argument, and eventually shown to have a direct algorithmic counterpart (strictly related to statistical inference). I will conclude by discussing current challenges in this field.
Short Biography:
Andrea Montanari received a Laurea degree in Physics in 1997 and a PhD in Theoretical Physics in 2001 (both from Scuola Normale Superiore in Pisa, Italy). He has been post-doctoral fellow at Laboratoire de Physique Theorique de l'Ecole Normale Superieure (LPTENS), Paris, France, and the Mathematical Sciences Research Institute, Berkeley, USA. Since 2002 he is Chargee de Recherche (a permanent research position with Centre National de la Recherche Scientifique, CNRS) at LPTENS. In 2006 he has been announced for the CNRS bronze medal in theoretical physics.