par Loris, Ignace 
Référence 28th European Conference on Operational Research, (28: 3-6/07/2016: Poznan, Poland)
Publication Non publié, 2016-07-06
Référence 28th European Conference on Operational Research, (28: 3-6/07/2016: Poznan, Poland)
Publication Non publié, 2016-07-06
Communication à un colloque
| Résumé : | A novel iterative algorithm for the solution of convex or non-convex optimization problems is presented. We assume that the objective function is the sum of a differentiable (possibly non-convex) function plus a convex (possibly non-differentiable) function. The algorithm is based on inexact knowledge of the proximal operator of the non-smooth part and uses a variable metric in combination with an Armijo line-search rule. In general we prove that all limit points of the iterates are stationary, while in the special case of a convex objective function we prove the convergence of the whole sequence to a minimizer, under the assumption that a minimizer exists. Moreover, verifiable criteria for the inexact computation of the proximal operator are given in some cases of practical interest. Finally, the algorithms are applied to various image reconstruction tasks.This presentation is based on joint work with S. Bonettini, F. Porta, M. Prato and S. Rebegoldi. |
