Journal of Convex Analysis 23 (2016), No. 2, 615--629
Copyright Heldermann Verlag 2016
Zero Duality Gap and Attainment with Possibly Non-Convex Data
Emil Ernst
Université Aix Marseille, Centrale Marseille, I2M -- UMR 7373, 13453 Marseille, France
Emil.Ernst@univ-amu.fr
Michel Volle
Université d'Avignon, 74 rue Louis Pasteur, 84029 Avignon Cedex 1, France
Michel.Volle@univ-avignon.fr
A newly defined notion of convex closedness regarding a set is used in order to state a necessary and sufficient criterion for the min-sup property in non necessarily convex primal-dual optimization problems, generalizing well-known theorems valid in the convex setting. Our main result is then applied to the classical penalty method.
Keywords: Dual optimization, min-sup property, convex closedness regarding a set, penalty method.
MSC: 49M30, 49N15, 52A20
[ Fulltext-pdf (158 KB)] for subscribers only.