New interval methods for constrained global optimization

• 成果类型: Article
• 署名作者: Markót, MC; Fernández, J; Casado, LG; Csendes, T
• 署名单位: Szeged University; Hungarian Academy of Sciences; European Space Agency; European Space Research & Technology Centre; University of Murcia; Universidad de Almeria; Szeged University
• 刊物名称: MATHEMATICAL PROGRAMMING
• ISSN/ISSBN: 0025-5610
• DOI: 10.1007/s10107-005-0607-2
• 发表日期: 2006
• 页码: 287-318
• 关键词: subdivision direction selection; bound methods; multisection; algorithm; criterion; facility; tests

摘要

Interval analysis is a powerful tool which allows to design branch-and-bound algorithms able to solve many global optimization problems. In this paper we present new adaptive multisection rules which enable the algorithm to choose the proper multisection type depending on simple heuristic decision rules. Moreover, for the selection of the next box to be subdivided, we investigate new criteria. Both the adaptive multisection and the subinterval selection rules seem to be specially suitable for being used in inequality constrained global optimization problems. The usefulness of these new techniques is shown by computational studies.