Mathematics and Statistics Colloquium
Friday (2/17/2006) at 2:30pm in 304 Pickard Hall
Dr. Dan Butnariu,
Department of Mathematics,
University of Haifa, Israel
"The proximal point algorithm and an augmented lagrangean method
In this talk we present a method of solving convex
optimization problems which also works in infinite dimensional spaces.
The method is based on a generalized proximal point algorithm. We show
that application of the generalized proximal point algorithm to the dual
optimization problem in conjunction with an iterative procedure of
updating the primal variables produces sequences of vectors (x^k,y^k) in
the primal-dual space which, in certain conditions, approximate
solutions of the Karush-Kuhn-Tucker system associated to the primal
problem and, implicitly, optimal pairs.