|
|
Vendredi 28 Novembre
Heure: |
11:00 - 12:30 |
Lieu: |
Salle B107, bâtiment B, Université de Villetaneuse |
Résumé: |
New applications of moment-SOS hierarchies |
Description: |
Victor Magron Semidefinite programming is relevant to a wide range of mathematic fields, including combinatorial optimization, control theory, matrix completion. In 2001, Lasserre introduced a hierarchy of semidefinite relaxations for particular polynomial instances of the Generalized Moment Problem (GMP). My talk emphasizes new applications of this moment-SOS hierarchy, investigated during my PhD and Postdoc research.
In the context of formal proofs for nonlinear optimization, one can combine the moment-SOS hierarchy with maxplus approximation of semiconvex functions. Such a framework is mandatory for formal certification of nonlinear inequalities, occurring by thousands in the proof of Kepler Conjecture by Hales.
I also present how to approximate, as closely as desired, the Pareto curve associated with bicriteria polynomial optimization problems or the projections of semialgebraic sets. For each problem, one builds a hierarchy of semidefinite programs, so that the sequence of bounds converges in L1 norm.
Finally, this hierarchy allows to analyze programs containing loop invariants with polynomial assignments. |
|
|