US EPA

4710736 

Journal Article 

THE MESH ADAPTIVE DIRECT SEARCH ALGORITHM WITH TREED GAUSSIAN PROCESS SURROGATES 

Gramacy, RB; Le Digabel, S 

2015 

11 

3 

419-447 

WOS:000360418500001 

This work introduces the use of the treed Gaussian processes (TGP) as a surrogate model within the mesh adaptive direct search (MADS) framework for constrained blackbox optimization. It extends the surrogate management framework (SMF) to nonsmooth optimization under general constraints. MADS uses TGP in two ways: one, as a surrogate for blackbox evaluations; and two, to evaluate statistical criteria such as the expected improvement and the average reduction in variance. The efficiency of the method is tested on five problems: a synthetic one with many local optima, a synthetic one implying a numerical method, one real application from a chemical engineering simulator for styrene production, one from contaminant cleanup and hydrology, and one multidisciplinary design optimization application. In all five cases we show that the TGP surrogate is preferable to a quadratic model and to MADS without any surrogate at all. 

blackbos optimization; treed Gaussian processes (TGP); mesh adaptive direct search (MADS); surrogate management framework (SMF)