Date
2014
Document Type
Thesis
Degree
Master of Science
Department
Industrial Engineering
First Adviser
Curtis, Frank E.
Abstract
An adaptive augmented Lagrangian algorithm is presented to overcome some undesirable behavior of traditional augmented Lagrangian methods. While the method has previously been proposed in \cite{AAL}, the goal in this thesis is to improve its practical performance. In particular, we propose an active set projected conjugate gradient (ASPCG) method for solving the subproblems of the adaptive augmented Lagrangian algorithm. The proposed ASPCG algorithm first estimates the optimal active set and then performs a projected conjugate gradient method to produce the exact or at least a good approximate solution updating the active set estimate when appropriate. We perform a series of numerical experiments to determine if the proposed algorithm is superior in some critical performance measures to the solver originally implemented in the adaptive augmented Lagrangian algorithm. In addition, we conduct experiments to monitor the performance of the adaptive augmented Lagrangian algorithm when some of its key features are modified.
Recommended Citation
Zhang, Wenda, "A Subproblem Algorithm for the Adaptive Augmented Lagrangian Method" (2014). Theses and Dissertations. 1692.
https://preserve.lehigh.edu/etd/1692