Home // FUTURE COMPUTING 2011, The Third International Conference on Future Computational Technologies and Applications // View article

A Computational Intelligence Algorithm for Simulation-driven Optimization Problems

Authors:
Yoel Tenne
Kazuhiro Izui
Shinji Nishiwaki

Keywords: expensive optimization problems, computational intelligence

Abstract:
Modern engineering design optimization often replaces laboratory experiments with computer simulations, resulting in what is commonly termed as expensive black-box optimization problems. In such problems, there will often exist candidate solutions which ‘crash’ the simulation and can thus lead to search stagnation and to a poor final result. Existing approaches to handle such solutions include discarding them altogether or assigning them a penalized fitness, but such approaches have significant demerits. Accordingly, this paper explores the fusion of a classifier into the optimization search to predict which solutions are expected to crash the simulation, and uses a modified objective function to bias the search towards valid ones, namely, which are expected not to crash the simulator. The further improve its performance, the proposed algorithm also continuously selects during the search an optimal type of classifier out of a family of candidates. To ensure the progress of the optimization search, it also employs a trust-region approach. Performance analysis using a representative real-world shape design optimization test case shows the efficacy of the proposed algorithm.

Pages: 127 to 134

Copyright: Copyright (c) IARIA, 2011

Publication date: September 25, 2011

Published in: conference

ISSN: 2308-3735

ISBN: 978-1-61208-154-0

Location: Rome, Italy

Dates: from September 25, 2011 to September 30, 2011