Numerical Optimization at Big-Level

Exploring the use of SQP methods for numerical optimization at a large scale

Abstract

Here we consider sequential quadraticprogramming methods (SQP methods) for the general optimization problem As with the BCL and LCL approaches, SQPmethods use a sequence of subproblems to _nd estimates  that reduce certain augmented Lagrangians. Animportant function of the subproblems is to estimate which inequalities in  are active. Rather many iterations (calledminor iterations) may be needed to solve each subproblem.