Argonne National Laboratory

Rejoinder: Modeling an Augmented Lagrangian for Blackbox Constrained Optimization

TitleRejoinder: Modeling an Augmented Lagrangian for Blackbox Constrained Optimization
Publication TypeJournal Article
Year of Publication2015
AuthorsGramacy, RB, Gray, GA, Le Digabel, S, Lee, HKH, Ranjan, P, Wells, G, Wild, SM
JournalTechnometrics
Volume58
Issue1
Pagination26-29
Other NumbersANL/MCS-P5401-0915
AbstractWe are grateful for the many insightful comments provided by the discussants. One team politely pointed out oversights in our literature review and the subsequent omission of a formidable comparator. Another made an important clarification about when a more aggressive variation (the so-called NoMax) would perform poorly. A third team offered enhancements to the framework, including a derivation of closed-form expressions and a more aggressive updating scheme; these enhancements were supported by an empirical study comparing new alternatives with old. The last team suggested hybridizing the statistical augmented Lagrangian (AL) method with modern stochastic search. Below we present our responses to these contributions and detail some improvements made to our own implementations in light of them. We conclude with some thoughts on statistical optimization using surrogate modeling and open-source software.  
DOI10.1080/00401706.2015.1106979
PDFhttp://www.mcs.anl.gov/papers/P5401-0915.pdf