Evolution of the Mads Algorithm by Developing Specific Features

Abstract: The topic of this talk is on Blackbox optimization (BBO), the study of applications, attributes, and solutions of optimization problems in which the values of one or more of the functions defining the problem are provided through blackboxes. The Mesh Adaptive Direct Search (Mads) is a derivative-free algorithm pioneered in 2006 for constrained BBO problems. This talk discusses recent advances to the Mads algorithm, including -i- the treatment of granular variables; -ii- dynamic scaling of variables; -iii- escaping unknown discontinuities; and -iv- revised convergence results for discontinuous functions.

Mathematics

Audience: researchers in the topic

PIMS-CORDS SFU Operations Research Seminar