The Exact Penalty Map for Nonsmooth and Nonconvex Optimization


Alfredo Noel Iusem
IMPA, Brazil

Friday, May 16, 2014
4:15 - 5:15 PM
Y2E2 105


Abstract:

Augmented Lagrangian duality provides zero duality gap and saddle point properties for nonconvex optimization. Hence, subgradient-like methods can be applied to the (convex) dual of the original problem, recovering the optimal value of the problem, but possibly not a primal solution. We prove that the recovery of a primal solution by such methods can be characterized in terms of the differentiability properties of the dual function, and the exact penalty properties of the primal-dual pair.

---

Operations Research Colloquia: http://or.stanford.edu/oras_seminars.html