Méthodes de faisceaux en programmation non convexe

Abstract

The aim of this work is to present bundle methods for solving convex and nonconvex programming problems. To this end, we begin by some indispensable recalls in nonsmooth optimization, then we present two bundle methods in convex optimization : a primal bundle method and a method based on the proximal trajectory. In a second part, we consider the non convex case by presenting two other bundle methods, one using the proximal trajectory and the other using the difference between two convex functions. Finally we report some numerical results to illustrate the behavior of the algorithm in the nonconvex case.

Date of Award	25 Jun 2004
Original language	French
Supervisor	Jean-Jacques STRODIOT (Supervisor), Van Hien Nguyen (Jury) & Geneviève Salmon (Jury)