Understanding search behavior via search landscape analysis in design optimization of optical structures

摘要

Search algorithms play a crucial role in systematic design optimization of optical structures. We demonstrate how search landscape analysis sheds light on the behavior of search algorithms in this context and thereby leads to valuable insight. Specifically, we present a case study in which search landscape analysis demystifies the surprisingly good performance of a simple stochastic local search algorithm-restarted iterative best improvement-on a challenging design optimization problem-combinatorial design optimization of a multi-spatial-mode photonic crystal waveguide bend that preserves modal content. Moreover, the search landscape analysis leads to insight potentially valuable in design optimization of a broad class of optical structures. (C) 2016 Optical Society of America